COT

Can Reasoning Help Large Language Models Capture Human Annotator Disagreement? (2026.eacl-long)

Challenge:	a lack of evaluation grounds for self-consistency on symbolic reasoning is unclear . however, it is unclear whether it improves performance on non-math questions involving encyclopedic knowledge.
Approach:	They establish a knowledge recall split for the popular MMLU benchmark by applying a data-driven heuristic from prior work.
Outcome:	The proposed knowledge recall split achieves an 89% accuracy on the MMLU benchmark.

Copied to clipboard

Jingwei Ni, Yu Fan, Vilém Zouhar, Donya Rooein, Alexander Miserlis Hoyle, Mrinmaya Sachan, Markus Leippold, Dirk Hovy, Elliott Ash

MathPrompter: Mathematical Reasoning using Large Language Models (2023.acl-industry)

Challenge:	Variation in human annotation (i.e., disagreements) is common in NLP, but it is unclear whether it is possible to model this variation in LLMs.
Approach:	They evaluate the influence of different reasoning settings on LLM disagreement modeling . RLVR-style reasoning degrades performance in disagreement modeling, they find .
Outcome:	The proposed reasoning settings improve LLM disagreement modeling, while RLVR-style reasoning degrades it.

Copied to clipboard

Shima Imani, Liang Du, Harsh Shrivastava

NumeroLogic: Number Encoding for Enhanced LLMs’ Numerical Reasoning (2024.emnlp-main)

Challenge:	Recent advances in natural language processing (NLP) can be attributed to massive scaling of Large Language Models (LLMs).
Approach:	They propose a technique that improves performance of Large Language Models (LLMs) on arithmetic problems along with increased reliance in the predictions.
Outcome:	The proposed technique improves performance on arithmetic problems and increases confidence in the output results.

Copied to clipboard

Eli Schwartz, Leshem Choshen, Joseph Shtok, Sivan Doveh, Leonid Karlinsky, Assaf Arbelle

Numerical Reasoning

Challenge:	Language models struggle with numerical and arithmetical tasks, such as multiplying 3-digit numbers.
Approach:	They propose a method to include the count of digits before each number instead of “42”.
Outcome:	The proposed format improves the reasoning process before generating the actual number.

DDPrompt: Differential Diversity Prompting in Large Language Models (2024.acl-short)

Copied to clipboard

Lin Mu, Wenhao Zhang, Yiwen Zhang, Peiquan Jin

Few shot chain-of-thought driven reasoning to prompt LLMs for open-ended medical question answering (2024.findings-emnlp)

Challenge:	Large Language Models (LLMs) have shown that their reasoning ability can be enhanced through approaches like Chain-of-Thought (CoT) prompting.
Approach:	They propose a method that generates differentially diverse reasoning paths for different types of questions by voting on the optimal prompts.
Outcome:	The proposed method improves LLMs' reasoning ability on complex reasoning tasks by learning from demonstrations while keeping their parameters frozen.

Copied to clipboard

Saeel Nachane, Ojas Gramopadhye, Prateek Chanda, Ganesh Ramakrishnan, Kshitij Jadhav, Yatin Nandwani, Dinesh Raghu, Sachindra Joshi

Disentangling the Effects of Unlearning in Measuring Parametric Faithfulness of Chain-of-Thought (2026.acl-srw)

Challenge:	Large Language models (LLMs) are increasingly utilized in the healthcare sector for query-related tasks.
Approach:	They propose a modified version of the MedQA-USMLE dataset, named MEDQA-OPEN, which contains open-ended medical questions without options to mimic clinical scenarios.
Outcome:	The proposed approach outperforms the state-of-the-art 5-shot CoT-based prompt by exploring multiple differential diagnoses and narrowing down to a final diagnosis using MCQ-ELIMINATIVE.

Copied to clipboard

Ryo Mitsuhashi, Gaku Morio, Ayana Niwa, Masahiro Kaneko, Kentaro Inui, Terufumi Morishita, Yuta Koreeda, Yasuhiro Sogawa

CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation (2025.emnlp-main)

Challenge:	Chain-of-Thought (CoT) has been debated as a model's faithfulness to internal reasoning process.
Approach:	They propose to use unlearning to measure parametric faithfulness of models by adjusting for unintended artifacts of unlearning.
Outcome:	The proposed metric accounts for the unintended artifacts of unlearning and shows that it is non-negligible.

Copied to clipboard

Zhenyi Shen, Hanqi Yan, Linhai Zhang, Zhanghao Hu, Yali Du, Yulan He

Enhancing User Safety: Context-Aware Detection of Offensive Query-Ad Pairs in Multimodal Search Advertising (2026.eacl-industry)

Challenge:	Prior implicit CoT methods have underperformed in terms of efficiency and robustness by relying on natural language tokens for reasoning.
Approach:	They propose a training framework that compresses natural language CoT into continuous space by aligning hidden states of a designated token.
Outcome:	The proposed framework outperforms the existing state-of-the-art in 3.1x compression rate and 28.2% accuracy on GSM8k scale.

Copied to clipboard

Gaurav Kumar, Qiangjian Xi, Tanmaya Shekhar Dabral, Hooshang Ghasemi, Abishek Krishnamoorthy, Danqing Fu, Rui Min, Emilio Antunez, Zhongli Ding, Pradyumna Narayana

DiaSynth: Synthetic Dialogue Generation Framework for Low Resource Dialogue Applications (2025.findings-naacl)

Challenge:	Multi-modal online advertisements require robust content moderation to ensure user safety . key challenges include nuanced, multi-modal nature of ads, severe data scarcity and class imbalance due to the rarity of offensive content .
Approach:	They propose a framework that detects offensive content only when a user's search query is paired with a specific ad .
Outcome:	The proposed framework reduces the serving of offensive query-ad pairs by more than 80% while maintaining the efficiency required for real-time advertising systems.

Copied to clipboard

Sathya Krishnan Suresh, Wu Mengjun, Tushar Pranav, EngSiong Chng

Text Generation

Challenge:	Existing research is limited by general or niche datasets that lack sufficient scale for training dialogue systems.
Approach:	They propose a synthetic dialogue generation framework that uses Large Language Models and Chain of Thought reasoning to generate dynamic, domain-specific dialogues with simulated personas and diverse conversational features.
Outcome:	The proposed framework outperforms existing frameworks on dialogue summarization and quality increases as the size of the LLM increases from 3B to 8B.

Recursion of Thought: A Divide-and-Conquer Approach to Multi-Context Reasoning with Language Models (2023.findings-acl)

Copied to clipboard

Soochan Lee, Gunhee Kim

Question-Analysis Prompting Improves LLM Performance in Reasoning Tasks (2024.acl-srw)

Challenge:	Existing methods to generate intermediate steps (CoT) are limited by the maximum context size due to various reasons.
Approach:	They propose a new inference framework that introduces several special tokens that the models can output to trigger context-related operations.
Outcome:	Extensive experiments with multiple architectures including GPT-3 show that the proposed framework significantly improves LMs’ inference capability.

Copied to clipboard

Dharunish Yugeswardeenoo, Kevin Zhu, Sean O’Brien

Can LLMs Learn From Mistakes? An Empirical Study on Reasoning Tasks (2024.findings-emnlp)

Challenge:	Existing methods to improve LLM performance have focused on sophisticating the model's step-by-step calculation.
Approach:	They propose a question analysis prompting strategy in which the model is prompted to explain the question in 'n' words before solving.
Outcome:	The proposed prompt outperforms state-of-the-art prompts on arithmetic and commonsense datasets and consistently ranks among the top-2 prompts.

Copied to clipboard

Shengnan An, Zexiong Ma, Siqi Cai, Zeqi Lin, Nanning Zheng, Jian-Guang Lou, Weizhu Chen

MENDER: Multi-hop Commonsense and Domain-specific CoT Reasoning for Knowledge-grounded Empathetic Counseling of Crime Victims (2025.naacl-srw)

Challenge:	Existing work has shown that simple learning can enhance the chain-of-thought (CoT) reasoning of large language models.
Approach:	They construct mistake-correction datasets to identify and correct mistakes in CoTs . they conclude that LLMs can learn from mistakes to enhance their CoT reasoning .
Outcome:	The proposed datasets show that LLMs can learn from mistakes to enhance their CoT reasoning performance.

Copied to clipboard

Abid Hossain, Priyanshu Priya, Armita Mani Tripathi, Pradeepika Verma, Asif Ekbal

LLMs on interactive feature collections with implicit dynamic decision strategy (2025.coling-main)

Challenge:	Experimental evaluations on counseling dialogue dataset, POEM validate MENDER’s efficacy in generating coherent, knowledge-grounded responses.
Approach:	They propose a multi-hop commonsensE and domaiN-specific Chain-of-Thought reasoning framework that integrates commonsense and domain knowledge via multi-hopping reasoning over the dialogue context.
Outcome:	Experimental evaluations on counseling dialogue dataset validate MENDER’s efficacy in generating coherent, empathetic, knowledge-grounded responses.

Copied to clipboard

Juyeon Heo, Vihari Piratla, Kyunghyun Lee, Hyonkeun Joh, Adrian Weller

Self-Harmonized Chain of Thought (2025.naacl-long)

Challenge:	Large Language Models (LLMs) struggle to efficiently narrow down the search space . external engineered systems may not fully utilize the inherent problem-solving capabilities of LLMs .
Approach:	They propose to implicitly guide Large Language Models to enhance their interactive feature collection abilities within a single prompt.
Outcome:	The proposed approach improves the performance of large language models in real-world scenarios.

Copied to clipboard

Ziqi Jin, Wei Lu

Think Like You Execute: Verifiable Chain of Thought from Program Traces (2026.acl-industry)

Challenge:	Existing methods for chain-of-thought prompting have limitations . arithmetic, commonsense, and symbolic reasoning tasks are challenging .
Approach:	They propose a method that unifies diverse solution paths into a consistent reasoning pattern.
Outcome:	The proposed method outperforms existing methods by 2.8% on reasoning tasks.

Copied to clipboard

Shailja Thakur, Vaibhav Saxena, Rohan Kulkarni, Shivdeep Singh, Parameswaran Selvam, Hiroshi Kanayama, Hima Patel

Automatic Model Selection with Large Language Models for Reasoning (2023.findings-emnlp)

Challenge:	Current synthetic Chain-of-Thought (CoT) training data often consists of plausible-sounding explanations generated by teacher models, not verifiable accounts of actual program behavior.
Approach:	They propose to ground CoT generation directly in program execution traces to improve reasoning capabilities.
Outcome:	The proposed pipeline improves performance on live code benchmarks and on cruxEval-output and cruxeval-input.

Copied to clipboard

James Zhao, Yuxi Xie, Kenji Kawaguchi, Junxian He, Michael Xie

LLMs Faithfully and Iteratively Compute Answers During CoT: A Systematic Analysis With Multi-step Arithmetics (2026.findings-eacl)

Challenge:	Chain-of-Thought and Program-Aided Language Models offer different strengths and weaknesses.
Approach:	They propose a model selection method that uses a large language model to select between two different reasoning methods.
Outcome:	The proposed method shows significant performance improvements across eight reasoning datasets with Codex, ChatGPT, and GPT-4.

Copied to clipboard

Keito Kudo, Yoichi Aoki, Tatsuki Kuribayashi, Shusaku Sone, Masaya Taniguchi, Ana Brassard, Keisuke Sakaguchi, Kentaro Inui

PromptSculptor: Multi-Agent Based Text-to-Image Prompt Optimization (2025.emnlp-demos)

Challenge:	Specifically, we examine when the LLMs’ answer is (pre)determined, especially before the CoT begins or after, and how strongly the information from CoT specifically has a causal effect on the final answer.
Approach:	They examine when the LLMs’ answer is (pre)determined, especially before the CoT begins or after, and how strongly the information from CoT specifically has a causal effect on the final answer.
Outcome:	The proposed model can generate reasoning chains while generating the reasoning chain on the fly.

Copied to clipboard

Dawei Xiang, Wenyan Xu, Kexin Chu, Tianqi Ding, Zixu Shen, Yiming Zeng, Jianchang Su, Wei Zhang

Arithmetic Reasoning with LLM: Prolog Generation & Permutation (2024.naacl-short)

Challenge:	PromptSculptor automates the iterative prompt optimization process for Text-to-Image models . previous work focused on generating detailed, high-quality prompts based on user feedback .
Approach:	They propose a framework that decomposes a task into four specialized agents . they use Chain-of-Thought reasoning to transform a short, vague user prompt into a comprehensive, refined prompt.
Outcome:	The proposed framework significantly improves output quality and reduces iterations needed for user satisfaction.

Copied to clipboard

Xiaocheng Yang, Bingsen Chen, Yik-Cheung Tam

Guided Knowledge Generation with Language Models for Commonsense Reasoning (2024.findings-emnlp)

Challenge:	Existing work has shown that large language models can generate arithmetic and commonsense reasoning, but they are not native to mathematical operations and symbolic manipulations.
Approach:	They propose to use large language models to generate Prolog programs to solve math problems using a code interpreter to generate arithmetic and symbolic formulas.
Outcome:	The proposed model outperforms CoT generation in the GSM8K benchmark across three LLMs.

Copied to clipboard

Xiao Wei, Haoran Chen, Hang Yu, Hao Fei, Qian Liu

Forest for the Trees: Overarching Prompting Evokes High-Level Reasoning in Large Language Models (2025.naacl-long)

Challenge:	Large Language Models (LLMs) have achieved notable success in commonsense reasoning tasks, benefiting from extensive world knowledge acquired through extensive pretraining.
Approach:	They propose a method to generate knowledge explanations and to automatically assign labels based on the probability of correct answers.
Outcome:	The proposed method outperforms baselines on four widely-used commonsense reasoning benchmarks and shows that it can generate high quality knowledge leading to correct answers.

Copied to clipboard

Haoran Liao, Shaohua Hu, Zhihao Zhu, Hao He, Yaohui Jin

Benchmarking and Mitigating the Impact of Noisy User Prompts in Medical VLMs via Cross-Modal Reflection (2026.eacl-industry)

Challenge:	Recent advances in large language models (LLMs) have greatly propelled the progress of natural language process (NLP).
Approach:	They propose a deductive paradigm that decomposes the reasoning process and a prompting method that elicits high-level thinking of large language models (LLMs).
Outcome:	The proposed method improves ChatGPT and CoT by 19.0% and 3.1% on MMLU’s College Physics, 8.8% and 2.3% on GSM8k, and 10.3% and 2.5% on StrategyQA, respectively.

Copied to clipboard

Zhiyu Xue, Reza Abbasi-Asl, Ramtin Pedarsani

Think Wider, Detect Sharper: Reinforced Reference Coverage for Document-Level Self-Contradiction Detection (2025.emnlp-main)

Challenge:	Existing medical vision-language models follow user-provided prompts blindly, a new study finds . current models are noisy, causing problems with reliability in real-world interactions .
Approach:	They propose a method to evaluate the influence of clinical prompts on medical vision-language models . they use cross-modal reflection chain-of-thought to train the model to produce reasoning paths .
Outcome:	The proposed method significantly improves the robustness against noisy prompts . existing Med-VLMs follow user-provided prompts blindly, the authors show .

Copied to clipboard

Yuhao Chen, Yuanjie Lyu, Shuochen Liu, Chao Zhang, Junhui Lv, Tong Xu

Chain-of-Thought Prompting Obscures Hallucination Cues in Large Language Models: An Empirical Evaluation (2025.findings-emnlp)

Challenge:	Recent approaches to document-level contradiction detection (DSCD) only gain marginal improvement and often introduce inconsistencies across repeated responses.
Approach:	They propose a method that combines supervised fine-tuning and reinforcement learning to enhance document-level contradiction detection (DSCD) they propose to use a task-specific reward function to expand the model’s reasoning scope, boosting both accuracy and consistency.
Outcome:	The proposed method significantly boosts Llama 3.1-8B-Instruct’s accuracy from 38.5% to 51.1%, and consistency from 59.6% to76.2%.

Copied to clipboard

Jiahao Cheng, Tiancheng Su, Jia Yuan, Guoxiu He, Jiawei Liu, Xinqi Tao, Jingwen Xie, Huaxia Li

AlphaFin: Benchmarking Financial Analysis with Retrieval-Augmented Stock-Chain Framework (2024.lrec-main)

Challenge:	Chain-of-Thought (CoT) prompting can mitigate hallucinations by encouraging step-by-step reasoning, but its impact on halluciation detection remains underexplored.
Approach:	They conduct an empirical evaluation of CoT prompting in Large Language Models (LLMs) to examine their impact on hallucination detection methods.
Outcome:	The proposed method significantly affects the internal states and token probability distributions of the LLM.

Copied to clipboard

Xiang Li, Zhenyu Li, Chen Shi, Yong Xu, Qing Du, Mingkui Tan, Jun Huang

Information Retrieval

Challenge:	Currently, ML&DL methods fail to provide reasons for stock trend predictions, lacking interpretability and reasoning processes. large language models (LLMs) suffer from hallucinations and are unable to keep up with the latest information.
Approach:	They develop a method to train large language models to handle financial analysis tasks . they use AlphaFin datasets to compare performance with traditional methods .
Outcome:	The proposed method improves stock trend prediction and financial question answering tasks.

Active Prompting with Chain-of-Thought for Large Language Models (2024.acl-long)

Copied to clipboard

Shizhe Diao, Pengcheng Wang, Yong Lin, Rui Pan, Xiang Liu, Tong Zhang

A Multi-modal Large Language Model with Graph-of-Thought for Effective Recommendation (2025.naacl-long)

Challenge:	Existing methods to annotate large language models rely on a fixed set of human-annotated exemplars, which are not always the most effective for different tasks.
Approach:	They propose a method to adapt large language models to different tasks with task-specific example prompts (annotated with human-designed CoT reasoning) they introduce several metrics to characterize uncertainty so as to select the most uncertain questions for annotation.
Outcome:	The proposed method significantly improves performance on eight complex reasoning tasks.

Copied to clipboard

Zixuan Yi, Iadh Ounis

Efficient Reasoning for LLMs through Speculative Chain-of-Thought (2026.findings-acl)

Challenge:	Large Language Models (LLMs) have demonstrated a remarkable capability in language understanding and text generation in various realworld scenarios.
Approach:	They propose a Graph-of-Thought prompting technique in a Multi-modal LLM to leverage the complex structure of interaction graphs.
Outcome:	The proposed model outperforms 12 existing state-of-the-art models on 6 benchmark datasets.

Copied to clipboard

Jikai Wang, Juntao Li, Jianye Hou, Yan Bowen, Lijun Wu, Min Zhang

Reasoning-Enhanced Domain-Adaptive Pretraining of Multimodal Large Language Models for Short Video Content Governance (2025.emnlp-industry)

Challenge:	Existing methods for efficient reasoning focus on reducing the number of model parameters or shortening the chain-of-thought length.
Approach:	They propose a speculative chain-of-thought (SCoT) method to reduce reasoning latency by accelerating average reasoning speed through large and small model collaboration.
Outcome:	The proposed method reduces reasoning latency by 48%66% and 21%49% on GSM8K, MATH, GaoKao, CollegeMath and Olympiad datasets.

Copied to clipboard

Zixuan Wang, Yu Sun, Hongwei Wang, Baoyu Jing, Xiang Shen, Xin Dong, Zhuolin Hao, Hongyu Xiong, Yang Song

Self-Explore: Enhancing Mathematical Reasoning in Language Models with Fine-grained Rewards (2024.findings-emnlp)

Challenge:	Existing approaches to identifying inappropriate content require extensive human-labeled data and lack cross-issue generalization.
Approach:	They propose a reasoning-enhanced multimodal large language model (MLLM) pretraining paradigm for unified inappropriate content detection.
Outcome:	The proposed model improves the MLLM's performance in both zero-shot and supervised fine-tuning settings and shows strong generalization capabilities to emergent, previously unseen issues.

Copied to clipboard

Hyeonbin Hwang, Doyoung Kim, Seungone Kim, Seonghyeon Ye, Minjoon Seo

Analyzing Chain-of-thought Prompting in Black-Box Large Language Models via Estimated V-information (2024.lrec-main)

Challenge:	Recent studies have shown that large language models can solve complex reasoning tasks with Chain-of-Thought Prompting.
Approach:	They propose a training method where the LLM is tasked to explore the first wrong step within the rationale and use such signals as fine-grained rewards for further improvement.
Outcome:	The proposed model improves on the GSM8K and MATH test sets by 11.57% and 2.89% on average compared to supervised fine-tuning (SFT).

Copied to clipboard

Zecheng Wang, Chunshan Li, Zhao Yang, Qingbin Liu, Yanchao Hao, Xi Chen, Dianhui Chu, Dianbo Sui

Improve Vision Language Model Chain-of-thought Reasoning (2025.acl-long)

Challenge:	Chain-of-Thought (CoT) prompting and large language models (LLMs) have shown great potential in improving performance on challenging reasoning tasks.
Approach:	They propose a new metric which extends the concept of pointwise V-information to black-box models and quantifies label-relevant new information introduced by CoT prompting.
Outcome:	The proposed metric extends the concept of pointwise V-information to black-box models, quantifying label-relevant new information introduced by CoT prompting beyond pre-existing label information.

Copied to clipboard

Ruohong Zhang, Bowen Zhang, Yanghao Li, Haotian Zhang, Zhiqing Sun, Zhe Gan, Yinfei Yang, Ruoming Pang, Yiming Yang

One Missing Piece for Open-Source Reasoning Models: A Dataset to Mitigate Cold-Starting Short CoT LLMs in RL (2025.acl-industry)

Challenge:	Current training recipes often rely on datasets dominated by short annotations with limited rationales, hindering the models' ability to generalize to tasks requiring comprehensive reasoning.
Approach:	They propose a two-stage post-training strategy that augments short answers with CoT reasoning generated by GPT-4o, enhancing the VLM's CoT capabilities through fine-tuning.
Outcome:	The proposed strategy enhances the model's CoT capabilities through fine-tuning and reinforcement learning.

Copied to clipboard

Hyungjoo Chae, Dongjin Kang, Jihyuk Kim, Beong-woo Kwak, Sunghyun Park, Haeju Park, Jinyoung Yeo, Moontae Lee, Kyungjae Lee

Enhancing Auto-regressive Chain-of-Thought through Loop-Aligned Reasoning (2026.eacl-long)

Challenge:	Existing large reasoning models are limited by their closed nature and high API costs and safety issues.
Approach:	They propose to build a long CoT dataset with existing short CoT LLMs that are not trained for inference-time scaling.
Outcome:	The proposed model achieves quality comparable to—or slightly below—R1 and is able to think longer and provide control over the thought budget to better manage the overthinking problem.

Copied to clipboard

Qifan Yu, Zhenyu He, Sijie Li, Zhou Xun, Jun Zhang, Jingjing Xu, Di He

ReList: A Multi-objective Reasoning Framework for Diversified Listwise Query Recommendation (2026.acl-industry)

Challenge:	Chain-of-Thought prompting is a powerful technique for enhancing language model’s reasoning capabilities, but generating long and correct CoT trajectories is challenging.
Approach:	They propose to align the steps of Chain-of-Thought reasoning with loop iterations and apply intermediate supervision during the training of Looped Transformers.
Outcome:	The proposed method generates accurate reasoning chains for complex problems exceeding training length, and improves performance of the auto-regressive model.

Copied to clipboard

Shuxian Bi, Chenxu Wang, Wenjie Wang, Yueqi Mou, Fuli Feng, Tang Biao, Peng Yan

What Makes Chain-of-Thought Prompting Effective? A Counterfactual Study (2023.findings-emnlp)

Challenge:	Existing methods for related search have limited semantic redundancy and wasted retrieval quota . generative retrieval approaches lack explicit reasoning, relying on superficial click-through rate rewards .
Approach:	They propose a framework that transforms related search into a reasoning-enhanced listwise generation task.
Outcome:	Experimental results show that ReList outperforms state-of-the-art methods in query diversity and user engagement.

Copied to clipboard

Aman Madaan, Katherine Hermann, Amir Yazdanbakhsh

Reasoning Implicit Sentiment with Chain-of-Thought Prompting (2023.acl-short)

Challenge:	Using a few-shot prompt, we examine the effects of symbols and patterns on in-context learning in large language models.
Approach:	They employ a counterfactual prompting approach by manipulating examples and testing the consequences on model behavior.
Outcome:	The proposed approach allows us to understand the relative contributions of symbols and patterns on in-context learning.

Copied to clipboard

Hao Fei, Bobo Li, Qian Liu, Lidong Bing, Fei Li, Tat-Seng Chua

Learning Multi-Step Reasoning by Solving Arithmetic Tasks (2023.acl-short)

Challenge:	In implicit sentiment analysis, the opinion cues come in an implicit and obscure manner.
Approach:	They propose a three-step prompting principle for THOR to step-by-step induce the implicit aspect, opinion and finally the sentiment polarity.
Outcome:	The proposed framework pushes the state-of-the-art (SoTA) by over 6% F1 on supervised setup and more strikingly, boosts the SoTA by over 50% F1 with THOR+GPT3.

Copied to clipboard

Tianduo Wang, Wei Lu

The Impact of Reasoning Step Length on Large Language Models (2024.findings-acl)

Challenge:	Recent studies have demonstrated large LMs’ impressive performance in solving math problems, but such ability seems only to emerge from models with abundant parameters.
Approach:	They propose to continuously pre-train LMs with the capabilities of multi-step reasoning by continuously pretraining them on a synthetic dataset MsAT.
Outcome:	The proposed method improves LMs' multi-step reasoning abilities on four math word problem datasets.

Copied to clipboard

Mingyu Jin, Qinkai Yu, Dong Shu, Haiyan Zhao, Wenyue Hua, Yanda Meng, Yongfeng Zhang, Mengnan Du

A Testset for Context-Aware LLM Translation in Korean-to-English Discourse Level Translation (2025.coling-main)

Challenge:	Long reasoning steps in LLMs improve reasoning abilities, but the correlation between their effectiveness and the length of reasoning steps remains largely unknown.
Approach:	They conducted experiments that expand and compress the rationale reasoning steps within CoT demonstrations while keeping all other factors constant.
Outcome:	The results show that lengthening the reasoning steps in prompts significantly enhances LLMs’ reasoning abilities across multiple datasets.

Copied to clipboard

Minjae Lee, Youngbin Noh, Seung Jin Lee

R3 Prompting: Review, Rephrase and Resolve for Chain-of-Thought Reasoning in Large Language Models under Noisy Context (2023.findings-emnlp)

Challenge:	Recent studies indicate that for high-resource languages, LLM surpasses encoder-decoder neural machine translation (NMT) models.
Approach:	They propose to construct a Korean-English discourse-level corpus with 600 text instances featuring six linguistic phenomena: lexical ambiguity, zero anaphora, slang, idiom, figurative language, and implicature.
Outcome:	The proposed corpus of 600 text instances features six linguistic phenomena, including lexical ambiguity, zero anaphora, slang, idiom, figurative language, and implicature.

Copied to clipboard

Qingyuan Tian, Hanlun Zhu, Lei Wang, Yang Li, Yunshi Lan

MolTC: Towards Molecular Relational Modeling In Language Models (2024.findings-acl)

Challenge:	Existing studies have evaluated LLMs under noise-free context but the dilemma for LLM to produce inaccurate results under noisy context has not been fully investigated.
Approach:	They propose a new method for CoT reasoning using Chain-of-Thought prompting that interacts with LLMs to perform key sentence extraction, variable declaration and answer prediction.
Outcome:	The proposed method outperforms existing CoT prompting methods on five reasoning tasks under noisy context.

Copied to clipboard

Junfeng Fang, Shuai Zhang, Chang Wu, Zhengyi Yang, Zhiyuan Liu, Sihang Li, Kun Wang, Wenjie Du, Xiang Wang

Prompt Space Optimizing Few-shot Reasoning Success with Large Language Models (2024.findings-naacl)

Challenge:	Molecular Relational Learning (MRL) is a promising way to understand interactions between molecular pairs.
Approach:	They propose a novel LLM-based multi-modal framework for molecular interaction modeling following Chain-of-Thought (CoT) theory which integrates graphical information of two molecules in pair.
Outcome:	The proposed framework integrates graphical information of two molecules in pair.

Copied to clipboard

Fobo Shi, Peijun Qing, Dong Yang, Nan Wang, Youbo Lei, Haonan Lu, Xiaodong Lin, Duantengchuan Li

LTRAG: Enhancing Autoformalization and Self-refinement for Logical Reasoning with Thought-Guided RAG (2025.findings-acl)

Challenge:	Prompt engineering is an essential technique for enhancing the abilities of large language models (LLMs) by providing explicit and specific instructions.
Approach:	They propose a new approach that uses text embeddings to obtain basis vectors by matrix decomposition and constructs a space for representing all prompts.
Outcome:	The proposed approach significantly outperforms state-of-the-art prompt paradigms on ten public reasoning benchmarks.

Copied to clipboard

Ruikang Hu, Shaoyu Lin, Yeliang Xiu, Yongmei Liu

MathMist: A Parallel Multilingual Benchmark Dataset for Mathematical Problem Solving and Reasoning (2026.findings-eacl)

Challenge:	Large language models (LLMs) have shown promise in natural language reasoning, especially with techniques like chain-of-thought prompting.
Approach:	They propose a framework to enhance autoformalization and self-refinement for logical reasoning with Retrieval-Augmented Generation (RAG) by building knowledge bases of thought-guided examples.
Outcome:	The proposed framework outperforms Logic-LM and LINC on FOLIO and AR-LSAT, and achieves an accuracy gain of 13% over Logic LM and the proposed methods on GPT-4 and AR LSAT.

Copied to clipboard

Mahbub E Sobhani, Md. Faiyaz Abdullah Sayeedi, Tasnim Mohiuddin, Md Mofijul Islam, Swakkhar Shatabda

Enhancing Healthcare LLM Trust with Atypical Presentations Recalibration (2024.findings-emnlp)

Challenge:	Existing benchmarks primarily focus on English or a narrow subset of high-resource languages, leaving significant gaps in assessing multilingual and cross-lingual mathematical reasoning.
Approach:	They propose a parallel multilingual benchmark for mathematical problem solving and reasoning that encompasses 2,890 parallel Bangla-English gold standard artifacts.
Outcome:	The proposed model encompasses 2,890 parallel Bangla-English gold standard artifacts, totaling 30K aligned question–answer pairs across thirteen languages, representing high-, medium-, and low-resource linguistic settings.

Copied to clipboard

Jeremy Qin, Bang Liu, Quoc Nguyen

Challenge:	Existing methods for eliciting and calibrating large language models have focused on general reasoning datasets, yielding only modest improvements.
Approach:	They propose a method which leverages atypical presentations to adjust model confidence estimates.
Outcome:	The proposed method reduces calibration errors by approximately 60% on three medical question answering datasets and outperforms existing methods such as vanilla verbalized confidence, CoT verbalised confidence and others.

PaD: Program-aided Distillation Can Teach Small Models Reasoning Better than Chain-of-thought Fine-tuning (2024.naacl-long)

Copied to clipboard

Xuekai Zhu, Biqing Qi, Kaiyan Zhang, Xinwei Long, Zhouhan Lin, Bowen Zhou

CiPO: Counterfactual Unlearning for Large Reasoning Models through Iterative Preference Optimization (2026.acl-long)

Challenge:	Large language models excel in various tasks, but their huge size and inaccessibility of parameters present challenges for practical deployment.
Approach:	They propose to use CoT data to distill task-specific ability from large language models to smaller models . they use reasoning programs to suppress errors in distilled data and improve distillation quality .
Outcome:	The proposed model outperforms LLMs on arithmetic reasoning, symbolic reasoning, and general ability.

Copied to clipboard

Junyi Li, Yongqiang Chen, Ningning Ding

ComfyUI-R1: Exploring Reasoning Models for Workflow Generation (2026.findings-acl)

Challenge:	Existing methods to unlearning large reasoning models do not remove unwanted knowledge from CoT traces or interfere with the reasoning process.
Approach:	They propose a framework that targets the CoT reasoning in Large Reasoning Models by generating a valid counterfactual reasoning trace for preference tuning.
Outcome:	Experiments on large LRMs show that CiPO completely removes knowledge from the intermediate CoT steps and the final answer while preserving the reasoning abilities of LRM.

Copied to clipboard

Zhenran Xu, Yiyu Wang, null Yangxue, Longyue Wang, Weihua Luo, Kaifu Zhang, Baotian Hu, Min Zhang

Plan-and-Solve Prompting: Improving Zero-Shot Chain-of-Thought Reasoning by Large Language Models (2023.acl-long)

Challenge:	ComfyUI-R1 is the first large reasoning model for automated workflow generation.
Approach:	They propose a large reasoning model for automated workflow generation that builds on curated knowledge bases and a two-stage framework to fine-tune models for cold start and reinforcement learning for incentivizing reasoning capability.
Outcome:	The proposed model achieves 97% format validity rate, high pass rate, node-level and graph-level F1 scores, surpassing prior state-of-the-art methods that employ leading closed-source models such as GPT-4o and Claude series.

Copied to clipboard

Lei Wang, Wanyu Xu, Yihuai Lan, Zhiqiang Hu, Yunshi Lan, Roy Ka-Wei Lee, Ee-Peng Lim

Towards Understanding Chain-of-Thought Prompting: An Empirical Study of What Matters (2023.acl-long)

Challenge:	Large language models (LLMs) have recently been shown to deliver impressive performance in various NLP tasks.
Approach:	They propose a plan-and-solve (PS) prompting that includes a few manual steps to generate reasoning steps and improves the quality of generated reasoning steps.
Outcome:	The proposed strategy outperforms Zero-shot-CoT on ten reasoning problems and has comparable performance to 8-shot CoT prompting on the math reasoning problem.

Copied to clipboard

Boshi Wang, Sewon Min, Xiang Deng, Jiaming Shen, You Wu, Luke Zettlemoyer, Huan Sun

Diagnosing Memorization in Chain-of-Thought Reasoning, One Token at a Time (2025.emnlp-main)

Challenge:	Chain-of-Thought (CoT) prompting can dramatically improve the multi-step reasoning abilities of large language models (LLMs).
Approach:	They propose to use Chain-of-Thought (CoT) prompting to encourage the LLM to generate intermediate rationales for solving a problem by providing a series of reasoning steps in the demonstrations.
Outcome:	The proposed model can generate coherent lines of reasoning even with invalid demonstrations while still generating coherent lines during inference.

Copied to clipboard

Huihan Li, You Chen, Siyuan Wang, Yixin He, Ninareh Mehrabi, Rahul Gupta, Xiang Ren

Analyzing LLM Instruction Optimization for Tabular Fact Verification (2026.findings-eacl)

Challenge:	Large Language Models (LLMs) perform well on reasoning benchmarks but often fail when inputs alter slightly, raising concerns about overreliance on memorization.
Approach:	They propose a framework for Source-aware Token-level Identification of Memorization which attributes each token in a reasoning chain to one of multiple memorization sources based on their statistical co-occurrence with the token in the pretraining corpus.
Outcome:	The proposed framework attributes each token in a reasoning chain to one of multiple memorization sources based on their statistical co-occurrence with the token in the pretraining corpus.

Copied to clipboard

Xiaotang Du, Giwon Hong, Wai-Chung Kwan, Rohit Saxena, Ivan Titov, Pasquale Minervini, Emily Allaway

On the Empirical Complexity of Reasoning and Planning in LLMs (2024.findings-emnlp)

Challenge:	evaluating instruction optimization for tabular fact verification is a key challenge for reliable NLP systems.
Approach:	They compare instruction optimization for tabular fact verification with a framework based on DSPy . they find that instruction optimization consistently improves verification accuracy .
Outcome:	The proposed method improves verification accuracy across four benchmarks and three model families.

Copied to clipboard

Liwei Kang, Zirui Zhao, David Hsu, Wee Lee

A Unified View on Emotion Representation in Large Language Models (2026.eacl-long)

Challenge:	Evidence shows that the relative performance of CoT, ToT, and their variants may vary from task to task.
Approach:	They propose to use chain-of-thought (CoT), tree-of thought (ToT), and related techniques to solve complex reasoning tasks with Large Language Models.
Outcome:	The proposed methods outperform the linear structure of CoT on hard reasoning tasks.

Copied to clipboard

Aishwarya Maheswaran, Maunendra Sankar Desarkar

Direct Evaluation of Chain-of-Thought in Multi-hop Reasoning with Knowledge Graphs (2024.findings-acl)

Challenge:	Recent studies show the presence of emotion concepts in the hidden state representations, but it’s unclear if the model has a robust representation consistent across different datasets.
Approach:	They propose a unified view to understand emotion representation in Large Language Models by experimenting with diverse datasets and prompts.
Outcome:	The proposed model can be interchanged between datasets with minimal impact on performance.

Copied to clipboard

Thi Nguyen, Linhao Luo, Fatemeh Shiri, Dinh Phung, Yuan-Fang Li, Thuy-Trang Vu, Gholamreza Haffari

Structured Data in NLP

Challenge:	Prior research on evaluating large language models focused on answer accuracy, neglecting the correctness of the generated CoT.
Approach:	They propose a discriminative and generative CoT evaluation paradigm to assess LLMs’ knowledge of reasoning and the accuracy of the generated CoT.
Outcome:	The proposed evaluation paradigm assesses LLMs’ knowledge of reasoning and the accuracy of the generated CoT.

CoT-RAG: Integrating Chain of Thought and Retrieval-Augmented Generation to Enhance Reasoning in Large Language Models (2025.findings-emnlp)

Copied to clipboard

Feiyang Li, Peng Fang, Zhan Shi, Arijit Khan, Fang Wang, Weihao Wang, null Zhangxin-hw, Cui Yongjian

Don’t Tell the Answer, Truly Guide the Reasoning During RL Rollouts (2026.findings-acl)

Challenge:	Chain-of-thought reasoning has two key limitations: lack of reliability when solely relying on LLM-generated reasoning chains and interference from natural language reasoning steps with the models’ inference logic.
Approach:	They propose a chain-of-thought reasoning framework with three key designs to address these issues.
Outcome:	The proposed framework improves the performance of large language models on complex tasks by incorporating knowledge graphs and learnable knowledge case-aware RAG.

Copied to clipboard

Xinyi Wang, Jinyi Han, Zishang Jiang, Tingyun li, Jiaqing Liang, Sihang Jiang, Zhaoqian Dai, Ma Shuguang, Fei Yu, Yanghua Xiao

Can LLMs Learn from Previous Mistakes? Investigating LLMs’ Errors to Boost for Reasoning (2024.acl-long)

Challenge:	Existing methods such as GRPO often break down when task difficulty exceeds the model’s capacity, resulting in sparse rewards and inefficient training.
Approach:	They propose to measure the compatibility between external guidance and a model's intrinsic policy by introducing an adaptive framework to enhance reasoning performance while explicitly preserving high Affinity.
Outcome:	The proposed framework outperforms baseline models while maintaining high Affinity.

Copied to clipboard

Yongqi Tong, Dawei Li, Sizhe Wang, Yujia Wang, Fei Teng, Jingbo Shang

Domaino1s: Guiding LLM Reasoning for Explainable Answers in High-Stakes Domains (2025.findings-acl)

Challenge:	Recent studies have shown the benefits to LLMs from fine-tuning golden-standard Chain-of-Thought rationales or using them as correct examples in few-shot prompting.
Approach:	They propose a new benchmark to test the effectiveness of large language models by leveraging errors to enhance reasoning capabilities.
Outcome:	The proposed methods can be used to fine-tune models in correct and incorrect domains, rather than tuning models to learn ground truth in traditional methods.

Copied to clipboard

Xu Chu, Zhijie Tan, Hanlin Xue, Guanyu Wang, Tong Mo, Weiping Li

ThinkAnswer Loss: Balancing Semantic Similarity and Exact Matching for LLM Reasoning Enhancement (2025.findings-emnlp)

Challenge:	Existing Large Language Models (LLMs) generate brief answers without reasoning processes and explanations.
Approach:	They propose supervised fine-tuning and tree search to enhance LLMs’ reasoning capabilities on domain tasks.
Outcome:	The proposed model improves on stock investment recommendation and legal reasoning QA tasks.

Copied to clipboard

Shan Yang, Kun Wu, Zeju Li, Linlin Zhang, Xiangyu Pei, Leike An, Yu Liu

Evaluating Step-by-Step Reasoning through Symbolic Verification (2024.findings-naacl)

Challenge:	Existing methods for knowledge distillation use Chain-of-Thought (CoT) and answer pairs, but they lack appropriate supervision signals.
Approach:	They propose a framework that decouples CoT and answer supervision . the framework applies semantic similarity constraints while maintaining strict literal matching for the answer .
Outcome:	The proposed framework decouples CoT and answer supervision while maintaining strict literal matching for the answer.

Copied to clipboard

YiFan Zhang, Hanlin Zhang, Li Li, Eric Xing

Unveiling Confirmation Bias in Chain-of-Thought Reasoning (2025.findings-acl)

Challenge:	Pre-trained language models (LMs) have shown remarkable reasoning performance using explanations or chain-of-thoughts (CoT)) for in-context learning.
Approach:	They propose to use symbolic examples to iteratively reason over symbolic examples and to recover Prolog’s backward chaining algorithm to iterate over KBs.
Outcome:	The proposed model performs better on length generalization benchmarks than CoT on explanations and chain-of-thoughts (CoT) tasks.

Copied to clipboard

Yue Wan, Xiaowei Jia, Xiang Lorraine Li

DRP: Distilled Reasoning Pruning with Mathematical Skill-aware Step Decomposition for Efficient Large Reasoning Models (2026.findings-acl)

Challenge:	Chain-of-thought (CoT) prompting has been widely adopted to enhance the reasoning capabilities of large language models (LLMs).
Approach:	They propose to examine how internal beliefs affect reasoning generation and reasoning-guided answer prediction in CoT by decomposing CoT into a two-stage process.
Outcome:	The proposed model beliefs affect reasoning generation and reasoning-guided answer prediction in CoT, and the results provide strong evidence of confirmation bias in LLMs.

Copied to clipboard

Yuxuan Jiang, Dawei Li, Francis Ferraro

Dual-Reasoner: Bridging Interleaved Atomicity and Streaming Latency via Thinking-while-Talking (2026.findings-acl)

Challenge:	Existing solutions to this problem are inference-time pruning and tuning-based distillation.
Approach:	They propose a framework that combines inference-time pruning with tuning-based distillation to enable efficient and accurate reasoning.
Outcome:	The proposed framework reduces token usage while improving accuracy on GSM8K and AIME tokens while avoiding performance drop.

Copied to clipboard

Yangzhuo Li, Shengpeng Ji, Yifu Chen, Tianle Liang, Haoyu Yang, null Junboli, Jun Fang, Lin Li, Qingyang Hong

LaRS: Latent Reasoning Skills for Chain-of-Thought Reasoning (2024.findings-emnlp)

Challenge:	Existing methods to integrate Chain-of-Thought into spoken dialogue models incur prohibitive latency.
Approach:	They propose a Streaming Masking Mechanism to ensure uninterrupted audio streaming . they use a quadruple-constraint system to reconstruct logical atomicity .
Outcome:	Experimental results show that Dual-Reasoner improves speech generation performance with low latency.

Copied to clipboard

Zifan Xu, Haozhu Wang, Dmitriy Bespalov, Xian Wu, Peter Stone, Yanjun Qi

Deciphering the Factors Influencing the Efficacy of Chain-of-Thought: Probability, Memorization, and Noisy Reasoning (2024.findings-emnlp)

Challenge:	Existing methods require human experts or pre-trained LLMs to describe the skill to guide the selection.
Approach:	They propose a new approach that uses unsupervised learning to create a latent space representation of rationales with a variable called a reasoning skill.
Outcome:	Empirical results show that LaRS outperforms SOTA skill-based selection methods . it processes example banks four times faster and reduces LLM inferences by half .

Copied to clipboard

Akshara Prabhakar, Thomas Griffiths, R. McCoy

What do Large Language Models Need for Machine Translation Evaluation? (2024.emnlp-main)

Challenge:	Chain-of-Thought (CoT) prompting has been shown to enhance the multi-step reasoning capabilities of Large Language Models (LLMs).
Approach:	They propose to use CoT prompting to analyze a symbolic reasoning task where letters are shifted forward some number of steps in the alphabet.
Outcome:	The proposed model performs well on a symbolic reasoning task, with three LLMs performing the task using CoT prompts.

Copied to clipboard

Shenbin Qian, Archchana Sindhujan, Minnie Kabra, Diptesh Kanojia, Constantin Orasan, Tharindu Ranasinghe, Fred Blain

PsyCoT: Psychological Questionnaire as Powerful Chain-of-Thought for Personality Detection (2023.findings-emnlp)

Challenge:	Existing research shows that large language models can perform better in machine translation tasks.
Approach:	They propose to use large language models for machine translation evaluations . authors explore what translation information is needed for LLMs to evaluate MT quality .
Outcome:	The proposed model performs comparable to fine-tuned multilingual pre-trained models.

Copied to clipboard

Tao Yang, Tianyuan Shi, Fanqi Wan, Xiaojun Quan, Qifan Wang, Bingzhe Wu, Jiaxiang Wu

INFORM : Information eNtropy based multi-step reasoning FOR large language Models (2023.emnlp-main)

Challenge:	Recent advances in large language models (LLMs) have demonstrated remarkable zero-shot performance across various NLP tasks.
Approach:	They propose a method which mimics the way individuals complete psychological questionnaires in a multi-turn dialogue manner and prompts an LLM to rate individual items at each turn.
Outcome:	The proposed method improves the performance and robustness of the standard GPT-3.5 personality detection task on two benchmark datasets.

Copied to clipboard

Chuyue Zhou, Wangjie You, Juntao Li, Jing Ye, Kehai Chen, Min Zhang

Controlling Out-of-Domain Gaps in LLMs for Genre Classification and Generated Text Detection (2025.coling-main)

Challenge:	Large language models (LLMs) have demonstrated exceptional performance with dedicated Chain-of-Thought (CoT) prompts.
Approach:	They propose a new method by introducing information entropy as a criteria on for CoT prompt selection.
Outcome:	The proposed model outperforms existing models on seven reasoning benchmarks using two language models.

Copied to clipboard

Dmitri Roussinov, Serge Sharoff, Nadezhda Puchnina

Information Retrieval

Challenge:	Recent advances in Large Language Models (LLMs) have pushed the boundaries of natural language processing, but their consistency is often limited when applied to unfamiliar domains.
Approach:	They propose a method that controls which predictive indicators are used and which are excluded during classification.
Outcome:	The proposed method reduces the OOD gap by up to 20 percentage points in a few-shot setup.

Benchmarking Uncertainty Metrics for LLM Target-Aware Search (2025.findings-emnlp)

Copied to clipboard

Pei-Fu Guo, Yun-Da Tsai, Shou-De Lin

Information Retrieval

Challenge:	Existing uncertainty metrics for LLM search methods do not capture the diverse types of uncertainty needed to guide different optimization goals.
Approach:	They propose a framework for uncertainty benchmarking that captures four different uncertainty types . the uncertainty types Answer, Correctness, Aleatoric, and Epistemic serve different optimization goals .
Outcome:	The proposed framework identifies four different uncertainty types . the uncertainty types serve different optimization goals in LLM search .

ReCoT-NER: Enhancing Zero-Shot Named Entity Recognition through Chain-of-Thought Prompting and Recall-Oriented Loss Optimization (2026.findings-acl)

Copied to clipboard

Dabin Fu, Fanghong Zhang

ACT-SQL: In-Context Learning for Text-to-SQL with Automatically-Generated Chain-of-Thought (2023.findings-emnlp)

Challenge:	Named Entity Recognition (NER) is a key component of natural language processing (NLP) but it is difficult to implement in specialized domains such as wind power fault diagnosis.
Approach:	They propose a reasoning-enhanced generative framework that integrates Chain-of-Thought prompting and recall-oriented loss optimization to address these challenges.
Outcome:	The proposed framework improves recall and overall F1 performance across general and industrial domains.

Copied to clipboard

Hanchong Zhang, Ruisheng Cao, Lu Chen, Hongshen Xu, Kai Yu

Just Ask One More Time! Self-Agreement Improves Reasoning of Language Models in (Almost) All Scenarios (2024.findings-acl)

Challenge:	Recent studies have focused on the development of semantic parsers within the framework of cross-domain analysis.
Approach:	They propose a method to generate auto-CoT exemplars using ACT-SQL and extend it to multi-turn text-to-Sql tasks.
Outcome:	The proposed method achieves SOTA performance on the Spider dev set among existing in-context learning approaches.

Copied to clipboard

Lei Lin, Jiayi Fu, Pengli Liu, Qingyang Li, Yan Gong, Junchen Wan, Fuzheng Zhang, Zhongyuan Wang, Di Zhang, Kun Gai

Recall with Reasoning: Chain-of-Thought Distillation for Mamba’s Long-Context Memory and Extrapolation (2025.emnlp-main)

Challenge:	chain-of-thought (CoT) prompting has been shown to be effective on complex reasoning tasks, but the naive greedy decoding used in CoT prompting causes the repetitiveness and local optimality.
Approach:	They propose a generalizable ensemble-optimization method that uses a set of reasoning paths to prompt a language model one more time to determine the optimal answer.
Outcome:	The proposed method can be generalized to almost all scenarios where the type of input questions and answer format of reasoning paths may be unknown.

Copied to clipboard

Jun-Yu Ma, Tianqing Fang, Zhisong Zhang, Hongming Zhang, Haitao Mi, Dong Yu

Text Generation

Challenge:	Existing long-context memory methods such as Mamba struggle with long-constituency when the length of the processed text exceeds the model's training length.
Approach:	They propose a method that uses chain-of-thought summarization to teach Mamba to actively recall and reason over long contexts.
Outcome:	Experiments on LONGMEMEVAL and HELMET show that RwR outperforms existing long-term memory methods while preserving short-context capabilities.

Prompting Few-shot Multi-hop Question Generation via Comprehending Type-aware Semantics (2024.findings-naacl)

Copied to clipboard

Zefeng Lin, Weidong Chen, Yan Song, Yongdong Zhang

NegotiationToM: A Benchmark for Stress-testing Machine Theory of Mind on Negotiation Surrounding (2024.findings-emnlp)

Challenge:	Existing approaches for multi-hop question generation rely on large annotated data . supervised approaches rely only on large labeled data, making it hard to perform tasks.
Approach:	They propose a type-aware semantics extraction-based chain-of-thought method for multi-hop question generation for documents . they first extract question types and essential semantic phrases from the given documents and the answer .
Outcome:	The proposed approach extracts question types and essential semantic phrases from documents and the answer.

Copied to clipboard

Chunkit Chan, Cheng Jiayang, Yauwai Yim, Zheye Deng, Wei Fan, Haoran Li, Xin Liu, Hongming Zhang, Weiqi Wang, Yangqiu Song

FinePrompt: Unveiling the Role of Finetuned Inductive Bias on Compositional Reasoning in GPT-4 (2023.findings-emnlp)

Challenge:	Theory of mind evaluations currently focus on testing models using machine-generated data or game settings prone to shortcuts and spurious correlations.
Approach:	They propose a benchmark to stress-test machine ToM in real-world negotiation surrounding covered multi-dimensional mental states.
Outcome:	The proposed benchmark builds upon the Belief-Desire-Intention theory and conducts the necessary empirical experiments to evaluate large language models.

Copied to clipboard

Jeonghwan Kim, Giwon Hong, Sung-Hyon Myaeng, Joyce Whang

Through the Valley: Path to Effective Long CoT Training for Small Language Models (2025.emnlp-main)

Challenge:	Large language models such as GPT-4 have demonstrated impressive capability to solve textual understanding problems at a level parallel to or surpassing state-of-the-art taskspecific models.
Approach:	They propose to transfer task-specific inductive biases from finetuned models to prompts to improve GPT-4's compositional reasoning capabilities.
Outcome:	The proposed prompt scheme shows competitive zero-shot and few-shot performances compared to existing prompts on complicated reasoning tasks.

Copied to clipboard

Renjie Luo, Jiaxi Li, Chen Huang, Wei Lu

Crowd Comparative Reasoning: Unlocking Comprehensive Evaluations for LLM-as-a-Judge (2025.acl-long)

Challenge:	Long chain-of-thought (CoT) supervision is effective for large language models . but small models trained on limited long CoT data experience performance degradation .
Approach:	They identify a phenomenon called Long CoT Degradation in small language models . long CoT data can be used to generate long chain-of-thought (CoT) responses .
Outcome:	The results show that models trained on 8k long CoT examples lose up to 75% of their original performance before fine-tuning.

Copied to clipboard

Qiyuan Zhang, Yufei Wang, Yuxin Jiang, Liangyou Li, Chuhan Wu, Yasheng Wang, Xin Jiang, Lifeng Shang, Ruiming Tang, Fuyuan Lyu, Chen Ma

Beyond the First Error: Process Reward Models for Reflective Mathematical Reasoning (2025.findings-emnlp)

Challenge:	Existing methods rely on majority voting or criteria expansion to capture detailed and detailed details, often leading to incomplete outcomes.
Approach:	They propose a method which introduces additional crowd responses to compare with the candidate responses, thereby exposing deeper and more comprehensive details within the candidate answers.
Outcome:	Experiments show that the proposed method improves evaluation reliability and achieves an average gain of 6.7% across five benchmarks.

Copied to clipboard

Zhaohui Yang, Chenghua He, Xiaowen Shi, Shihong Deng, Linjing Li, Qiyue Yin, Daxin Jiang

Re-TASK: Revisiting LLM Tasks from Capability, Skill, and Knowledge Perspectives (2025.findings-acl)

Challenge:	Existing methods for training effective PRMs focus on the first incorrect step and all preceding steps, assuming that all subsequent steps are incorrect.
Approach:	They propose a data annotation method specifically designed to score the long CoT reasoning process by using an LLM-based judger for annotation.
Outcome:	The proposed method improves PRMs' ability to identify effective self-correction behaviors and reasoning based on erroneous steps.

Copied to clipboard

Zhihu Wang, Shiwan Zhao, Yu Wang, Heyuan Huang, Sitao Xie, Yubo Zhang, Jiaxin Shi, Zhixing Wang, Hongyan Li, Junchi Yan

Challenge:	Existing approaches to solving complex tasks with large language models (LLMs) fail to decompose tasks accurately or execute subtasks effectively.
Approach:	They propose a Chain-of-Learning (CoL) paradigm that highlights task dependencies on specific capability items, further broken down into their constituent knowledge and skill components.
Outcome:	The proposed model improves Yi-1.5-9B and Llama3-Chinese-8B for legal tasks by 45.00% and 24.50% on different domains.

The Art of SOCRATIC QUESTIONING: Recursive Thinking with Large Language Models (2023.emnlp-main)

Copied to clipboard

Jingyuan Qi, Zhiyang Xu, Ying Shen, Minqian Liu, Di Jin, Qifan Wang, Lifu Huang

Mitigating Visual Forgetting via Take-along Visual Conditioning for Multi-modal Long CoT Reasoning (2025.acl-long)

Challenge:	Chain-of-Thought (CoT) prompting relies on the initial decisions, causing errors in early steps to accumulate and impact the final answers.
Approach:	They propose a divide-and-conquer style algorithm that leverages large language models to raise and answer sub-questions until collecting enough information to tackle the original one.
Outcome:	The proposed algorithm is more robust to errors and errors than CoT prompting and Tree-of-Thought prompting methods.

Copied to clipboard

Hai-Long Sun, Zhun Sun, Houwen Peng, Han-Jia Ye

WavLLM: Towards Robust and Adaptive Speech Large Language Model (2024.findings-emnlp)

Challenge:	Recent advances in Large Language Models (LLMs) have demonstrated enhanced reasoning capabilities, evolving from simple Chain-of-Thought (CoT) prompting to advanced, product-oriented solutions like OpenAI o1 .
Approach:	They propose a strategy that shifts image input to critical reasoning stages and compresses redundant visual tokens via dynamic pruning.
Outcome:	The proposed model achieves state-of-the-art on five mathematical reasoning benchmarks (+3.4% vs previous sota) and demonstrates iterative reasoning capabilities for complex multi-step tasks.

Copied to clipboard

Shujie Hu, Long Zhou, Shujie Liu, Sanyuan Chen, Lingwei Meng, Hongkun Hao, Jing Pan, Xunying Liu, Jinyu Li, Sunit Sivasankaran, Linquan Liu, Furu Wei

How Do Humans Write Code? Large Models Do It the Same Way Too (2024.emnlp-main)

Challenge:	Recent advances in large language models (LLMs) have expanded their scope to encompass multimodal functions.
Approach:	They propose a robust and adaptive speech large language model with dual encoders . they validate the model on universal speech benchmarks and apply it to specialized speech-question-answer datasets based on a CoT approach .
Outcome:	The proposed model achieves state-of-the-art performance across a range of speech tasks on the same model size.

Copied to clipboard

Long Li, Xuzheng He, Haozhe Wang, Linlin Wang, Liang He

ECHo: A Visio-Linguistic Dataset for Event Causality Inference via Human-Centric Reasoning (2023.findings-emnlp)

Challenge:	Program-of-Thought (PoT) replaces natural language-based Chain-ofThough (CoT) but introduces more reasoning errors, such as incorrect formulas or flawed logic, compared to CoT.
Approach:	They propose a method that integrates CoT and Program-of-Thought to achieve more accurate reasoning and reinforcement learning.
Outcome:	The proposed method achieves an average improvement of 6.5% on the Llama-Base model and 4.3% on the Mistral-Bass model across 8 mathematical calculation datasets.

Copied to clipboard

Yuxi Xie, Guanzhen Li, Min-Yen Kan

Human and LLM-Based Resume Matching: An Observational Study (2025.findings-naacl)

Challenge:	ECHo is a diagnostic dataset of event causality inference grounded in visio-linguistic social scenarios.
Approach:	They propose a diagnostic dataset of event causality inference grounded in visio-linguistic social scenarios.
Outcome:	The proposed framework examines the reasoning capability of current AI systems on three human-centric tasks.

Copied to clipboard

Swanand Vaishampayan, Hunter Leary, Yoseph Berhanu Alebachew, Louis Hickman, Brent A. Stevenor, Weston Beck, Chris Brown

Prompted LLMs as Chatbot Modules for Long Open-domain Conversation (2023.findings-acl)

Challenge:	Resume matching assesses the extent to which candidates qualify for jobs based on the content of resumes.
Approach:	They compare GPT-4 and human ratings for resumes submitted to job openings from diverse fields using real-world evaluation criteria.
Outcome:	The proposed model improves the quality of LLM ratings and does not show bias.

Copied to clipboard

Gibbeum Lee, Volker Hartmann, Jongho Park, Dimitris Papailiopoulos, Kangwook Lee

Dialogue Systems & Conversational Agents

Challenge:	Using pre-trained large language models (LLMs) as individual modules for long-term consistency and flexibility is a challenge for open-domain chatbots due to the computational burden of updating models with billions of parameters and the scarcity of data in the dialogue domain.
Approach:	They propose a method that uses pre-trained large language models as individual modules for long-term consistency and flexibility.
Outcome:	The proposed method is on par with fine-tuned chatbot models in open-domain conversations, showing it can create consistent and engaging chatbots.

ThinkSwitcher: When to Think Hard, When to Think Fast (2025.findings-emnlp)

Copied to clipboard

Guosheng Liang, Longguang Zhong, Ziyi Yang, Xiaojun Quan

CSEval: Towards Automated, Multi-Dimensional, and Reference-Free Counterspeech Evaluation using Auto-Calibrated LLMs (2025.naacl-long)

Challenge:	Large reasoning models excel at solving complex tasks by leveraging long chain-of-thought (CoT) reasoning.
Approach:	They propose a framework that enables a single LRM to dynamically switch between short and long CoT modes based on task complexity.
Outcome:	The proposed framework reduces computational cost by 20-30% while maintaining high accuracy on complex tasks.

Copied to clipboard

Amey Hengle, Aswini Kumar Padhi, Anil Bandhakavi, Tanmoy Chakraborty

Multimodal Causal Reasoning Benchmark: Challenging Multimodal Large Language Models to Discern Causal Links Across Modalities (2025.findings-acl)

Challenge:	Current evaluation methods do not capture complex attributes of counterspeech quality, such as contextual relevance, aggressiveness, or argumentative coherence.
Approach:	They propose to use a dataset and framework to evaluate counterspeech quality across four dimensions: contextual relevance, aggressiveness, argument-coherence, and suitability.
Outcome:	The proposed method outperforms ROUGE, METEOR, and BertScore in correlating with human judgement, indicating a significant improvement in automated counterspeech evaluation.

Copied to clipboard

Zhiyuan Li, Heng Wang, Dongnan Liu, Chaoyi Zhang, Ao Ma, Jieting Long, Weidong Cai

LLM-Driven Implicit Target Augmentation and Fine-Grained Contextual Modeling for Zero-Shot and Few-Shot Stance Detection (2025.emnlp-main)

Challenge:	Existing MLLMs lack robustness in multimodal causal reasoning compared to their performance in textual settings.
Approach:	They propose a novel multimodal chain-of-thought (CoT) reasoning benchmark that leverages siamese images and text pairs to challenge MLLMs.
Outcome:	The proposed benchmark leverages siamese images and text pairs to challenge MLLMs.

Copied to clipboard

Yanxu Ji, Jinzhong Ning, Yijia Zhang, Zhi Liu, Hongfei Lin

Sentiment Analysis

Challenge:	Recent studies on zero-shot and few-shot stance detection neglect implicit yet semantically important targets.
Approach:	They propose a framework that uses Large Language Models to annotate implicit targets . they also propose 'DyMCA' to dynamically adjust text-target contributions based on context .
Outcome:	The proposed framework achieves state-of-the-art on a benchmark dataset.

NeuroSym-Cal: Bridging the Reasoning-Execution Gap in Code Generation via Hierarchical Calibration (2026.findings-acl)

Copied to clipboard

Peiyang Liu, Yining Wang, Youru Li, Long Li, Zhi Cai, Wei Ye

Chain-of-Rank: Enhancing Large Language Models for Domain-Specific RAG in Edge Device (2025.findings-naacl)

Challenge:	Existing calibration methods rely on the assumption that consensus implies correctness . Existing methods fail under systematic errors, leading to miscalibrated high-confidence predictions.
Approach:	They propose a hierarchical calibration framework that measures confidence at two levels . they propose sensitivity analysis to measure local curvature of deductive process .
Outcome:	The proposed framework de-saturates overconfident errors and improves selective generation performance on OOD benchmarks.

Copied to clipboard

Juntae Lee, Jihwan Bang, Kyuhong Shim, Seunghan Yang, Simyung Chang

GCoT-Decoding: Unlocking Deep Reasoning Paths for Universal Question Answering (2026.findings-acl)

Challenge:	Retrieval-augmented generation (RAG) is valuable in specialized domains where precision is critical.
Approach:	They propose a chain-of-rank algorithm which allows LLMs to access a target domain early via finetuning.
Outcome:	The proposed method achieves state-of-the-art in benchmarks and analyzes its efficacy.

Copied to clipboard

Guanran Luo, Wentao Qiu, Zhongquan Jian, Meihong Wang, Qingqiang Wu

Verify-and-Edit: A Knowledge-Enhanced Chain-of-Thought Framework (2023.acl-long)

Challenge:	Recent work on Chain-of-Thought reasoning requires manual prompts to guide the model.
Approach:	They propose a general decoding strategy that generates CoT-style reasoning paths without prompts.
Outcome:	The proposed method maintains strong performance on fixed and free QA tasks and achieves significant improvements on free qa.

Copied to clipboard

Ruochen Zhao, Xingxuan Li, Shafiq Joty, Chengwei Qin, Lidong Bing

SeLaR: Selective Latent Reasoning in Large Language Models (2026.acl-long)

Challenge:	Large language models (LLMs) have a number of shortcomings, including lack of factual correctness.
Approach:	They propose a framework to increase prediction factuality by post-editing reasoning chains . they propose to use large language models to generate interpretable reasoning chains.
Outcome:	The proposed framework leads to accuracy improvements in open-domain question-answering tasks.

Copied to clipboard

Renyu Fu, Guibo Luo

MixKVQ: Query-Aware Mixed-Precision KV Cache Quantization for Long-Context Reasoning (2026.acl-long)

Challenge:	Recent latent reasoning approaches replace discrete tokens with soft embeddings or hidden states, but they often suffer from two issues: (1) global activation injects perturbations into high-confidence steps, impairing reasoning stability; and (2) soft embeds quickly collapse toward the highest-probability token, limiting exploration of alternative trajectories.
Approach:	They propose a lightweight and training-free framework that replaces discrete tokens with soft embeddings or hidden states to address these challenges.
Outcome:	Experiments on five reasoning benchmarks show that SeLaR outperforms standard CoT and state-of-the-art training-free methods.

Copied to clipboard

Tao Zhang, Ziqian Zeng, Hao Peng, Huiping Zhuang, Cen Chen

Large Language Model for Multi-Domain Translation: Benchmarking and Domain CoT Fine-tuning (2024.findings-emnlp)

Challenge:	Existing low-bit quantization methods often exhibit severe performance degradation on complex reasoning tasks.
Approach:	They propose a plug-and-play method that uses a key channel's intrinsic quantization difficulty and relevance to the query to identify and preserve critical key channels that need higher precision.
Outcome:	Experiments on complex reasoning datasets show that the proposed method outperforms low-bit methods at a substantially reduced memory footprint.

Copied to clipboard

Tianxiang Hu, Pei Zhang, Baosong Yang, Jun Xie, Derek Wong, Rui Wang

On Measuring Faithfulness or Self-consistency of Natural Language Explanations (2024.acl-long)

Challenge:	Achieving consistent high-quality machine translation across diverse domains remains a challenge due to limited and imbalanced parallel training data available in various domains.
Approach:	They propose a domain Chain of Thought technique that uses the multi-domain intelligence of LLMs to improve translation performance.
Outcome:	The proposed method achieves significant improvements in translation accuracy and domain robustness over traditional fine-tuning on a small dataset of four domains.

Copied to clipboard

Letitia Parcalabescu, Anette Frank

Explainability & Interpretability in NLP

Challenge:	Large language models (LLMs) can explain their predictions through post-hoc or Chain-of-Thought explanations, but they can also make unfaithful explanations that hide their sensitivity to biasing inputs.
Approach:	They propose to use a model-based consistency test to judge the faithfulness of post-hoc or Chain-of-Thought explanations rather than model-specific faithfulness tests.
Outcome:	The proposed tests do not measure faithfulness to model explainability but rather their self-consistency at output level.

Teaching Small Language Models Reasoning through Counterfactual Distillation (2024.emnlp-main)

Copied to clipboard

Tao Feng, Yicheng Li, Li Chenglin, Hao Chen, Fei Yu, Yin Zhang

Large Language Models as an Indirect Reasoner: Contrapositive and Contradiction for Automated Reasoning (2025.coling-main)

Challenge:	Large language models (LLMs) have demonstrated remarkable performance in a wide range of downstream tasks.
Approach:	They propose a counterfactual distillation framework that leverages LLMs to generate high-quality counterfacts and utilizes multi-view CoT to enhance the diversity of reasoning samples.
Outcome:	The proposed framework enhances reasoning capabilities of large language models and is more robust to OOD data.

Copied to clipboard

Yanfang Zhang, Yiliu Sun, Yibing Zhan, Dapeng Tao, Dacheng Tao, Chen Gong

Does Chain-of-Thought Reasoning Really Reduce Harmfulness from Jailbreaking? (2025.findings-acl)

Challenge:	Recent studies have focused on improving the ability of Large Language Models to perform complex reasoning.
Approach:	They propose a Direct-Indirect Reasoning method that integrates DR and IR as parallel reasoning paths that are merged to derive the final answer.
Outcome:	The proposed method outperforms existing methods on four datasets related to logical reasoning and proof.

Copied to clipboard

Chengda Lu, Xiaoyu Fan, Yu Huang, Rongwu Xu, Jijie Li, Wei Xu

ChartVerse: Scaling Chart Reasoning via Reliable Programmatic Synthesis from Scratch (2026.acl-long)

Challenge:	Existing jailbreak attacks fail against reasoning models enhanced by Chain-of-Thought (CoT) reasoning.
Approach:	They propose a jailbreak method that uses Chain-of-Thought reasoning to reduce harmfulness from jailbreaking.
Outcome:	The proposed jailbreak method performs well against open AI models and deepseek-R1 reasoning models.

Copied to clipboard

Zheng Liu, Honglin Lin, Xiaoyang Wang, Xin Gao, Yu Li, Mengzhang Cai, Yun Zhu, Zhanping Zhong, Qizhi Pei, Zhuoshi Pan, Xiaoran Shang, Conghui He, Bin Cui, Wentao Zhang, Lijun Wu

Bridging the Dynamic Perception Gap: Training-Free Draft Chain-of-Thought for Dynamic Multimodal Spatial Reasoning (2025.findings-emnlp)

Challenge:	Existing open-source vision language models lack high-quality training data for chart reasoning . current models are simplistic and repetitive, while associated QA pairs are prone to hallucinations .
Approach:	They propose a framework to synthesize complex charts and reliable reasoning data from scratch.
Outcome:	Experimental results show that ChartVerse-8B surpasses existing models in QA and difficulty . lack of high-quality training data hampers development of open-source models .

Copied to clipboard

Siqu Ou, Hongcheng Liu, Pingjie Wang, Yusheng Liao, Chuan Xuan, Yanfeng Wang, Yu Wang

DICE: Structured Reasoning in LLMs through SLM-Guided Chain-of-Thought Correction (2025.emnlp-main)

Challenge:	Existing methods for dynamic spatial reasoning are limited to text or static visual domains .
Approach:	They propose a framework that augments textual reasoning chains with dynamic visual drafts .
Outcome:	The proposed framework outperforms existing methods in dynamic spatial reasoning tasks.

Copied to clipboard

Yiqi Li, Yusheng Liao, Zhe Chen, Yanfeng Wang, Yu Wang

SPPO: Sequence-Level PPO for Long-Horizon Reasoning Tasks (2026.acl-long)

Challenge:	Large language models (LLMs) often prioritize reasoning over adherence to detailed instructions due to high computational costs and limited parameter access.
Approach:	They propose a lightweight framework that guides small language models to refine LLMs’ outputs through chain-of-thought correction.
Outcome:	The proposed framework improves the average format accuracy and content correctness of LLM outputs by 35.4% and 29.4%, respectively, achieving state-of-the-art (SOTA) performance over other competitive baselines.

Copied to clipboard

Tianyi Wang, Yixia Li, Long Li, Yibiao Chen, Shaohan Huang, Yun Chen, Peng Li, Yang Liu, Guanhua Chen

Unleashing Multi-Hop Reasoning Potential in Large Language Models through Repetition of Misordered Context (2025.findings-naacl)

Challenge:	Proximal Policy Optimization (PPO) is central to aligning Large Language Models with verifiable rewards.
Approach:	They propose a scalable algorithm that harmonizes sample efficiency with stability of outcome-based updates.
Outcome:	The proposed algorithm outperforms standard PPO and matches the performance of computation-heavy group-based methods.

Copied to clipboard

Sangwon Yu, Ik-hwan Kim, Jongyoon Song, Saehyung Lee, Junsung Park, Sungroh Yoon

Answering Questions by Meta-Reasoning over Multiple Chains of Thought (2023.emnlp-main)

Challenge:	Multi-hop reasoning requires multi-step reasoning based on supporting documents within a given context.
Approach:	They propose a method that prompts the model by repeatedly presenting the context.
Outcome:	The proposed method improves the F1 score by 30%p on multi-hop QA tasks and increases accuracy by 70%p on a synthetic task.

Copied to clipboard

Ori Yoran, Tomer Wolfson, Ben Bogin, Uri Katz, Daniel Deutch, Jonathan Berant

Question Answering

Challenge:	Modern systems for multi-hop question answering (QA) break questions into a sequence of reasoning steps, termed chain-of-thought (CoT) Often, multiple chains are sampled and aggregated, but the intermediate steps themselves are discarded.
Approach:	They propose a method which prompts large language models to meta-reason over multiple chains of thought rather than aggregate their answers.
Outcome:	The proposed approach outperforms baselines on 7 multi-hop QA datasets.

VisualCoder: Guiding Large Language Models in Code Execution with Fine-grained Multimodal Chain-of-Thought Reasoning (2025.findings-naacl)

Copied to clipboard

Cuong Le Chi, Chau Truong Vinh Hoang, Phan Nhật Huy, Dung D. Le, Tien N Nguyen, Nghi D. Q. Bui

Enhancing Reasoning Abilities of Small LLMs with Cognitive Alignment (2025.emnlp-main)

Challenge:	Existing approaches to enhance large language models' ability to predict program behavior struggle with dynamic reasoning tasks.
Approach:	They propose a visual control flow graph that integrates CoT reasoning with a control flow . they aim to improve performance in program behavior prediction, error detection and output generation .
Outcome:	The proposed approach improves program behavior prediction, error detection, and output generation.

Copied to clipboard

Wenrui Cai, Chengyu Wang, Junbing Yan, Jun Huang, Xiangzhong Fang

GAMEBoT: Transparent Assessment of LLM Reasoning in Games (2025.acl-long)

Challenge:	Existing methods to distill chain-of-thought (CoT) results from large language reasoning models (LRMs) to small models are ineffective and require substantial amount of annotated data.
Approach:	They propose a Critique-Rethink-Verify system for training small language reasoning models that can be critiquized according to the cognitive capabilities of smaller models.
Outcome:	The proposed system outperforms other methods on challenging reasoning benchmarks.

Copied to clipboard

Wenye Lin, Jonathan Roberts, Yunhan Yang, Samuel Albanie, Zongqing Lu, Kai Han

Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Do (2026.acl-long)

Challenge:	Existing efforts to create benchmarks that move beyond superficial pattern recognition to delve into the profound reasoning skills required for problemsolving face challenges such as insufficient interpretability, performance saturation or data contamination.
Approach:	They propose a gaming arena designed for rigorous assessment of LLM reasoning capabilities.
Outcome:	The proposed framework decomposes complex reasoning into predefined modular subproblems and generates ground truth for these subproblem types.

Copied to clipboard

Zhuoran Jin, Kejian Zhu, Hongbang Yuan, Yupu Hao, Pengfei Cao, Yubo Chen, Kang Liu, Jun Zhao

An Investigation of Neuron Activation as a Unified Lens to Explain Chain-of-Thought Eliciting Arithmetic Reasoning of LLMs (2024.acl-long)

Challenge:	Existing open-source models often yield only marginal overall improvements, possibly due to an overemphasis on mathematical reasoning at the expense of broader capabilities.
Approach:	They evaluate 12 multimodal tasks using 14 non-reasoning models and 8 reasoning models.
Outcome:	The proposed method is effective in multimodal reasoning tasks, the authors show . they show that it lacks the ability to maintain deep visual introspection throughout the reasoning process.

Copied to clipboard

Daking Rai, Ziyu Yao

Wrong-of-Thought: An Integrated Reasoning Framework with Multi-Perspective Verification and Wrong Information (2024.findings-emnlp)

Challenge:	Prior work has focused on ablating components in the CoT prompt, but the reason why these components are important to LLM reasoning is not explored.
Approach:	They investigate "neuron activation" as a lens to provide a unified explanation to previous work . they propose an approach to automatically identify neurons that imply arithmetic reasoning .
Outcome:	The proposed approach can explain the importance of components in a CoT prompt . it also automatically identifies neurons that imply arithmetic reasoning .

Copied to clipboard

Yongheng Zhang, Qiguang Chen, Jingxuan Zhou, Peng Wang, Jiasheng Si, Jin Wang, Wenpeng Lu, Libo Qin

WebAgent-R1: Training Web Agents via End-to-End Multi-Turn Reinforcement Learning (2025.emnlp-main)

Challenge:	Chain-of-Thought (CoT) is a key technique for enhancing the performance of Large Language Models.
Approach:	They propose a framework that optimizes outputs by utilizing wrong information and multi-perspective verification.
Outcome:	The proposed framework surpasses all baselines on 8 datasets and 5 LLMs.

Copied to clipboard

Zhepei Wei, Wenlin Yao, Yao Liu, Weizhi Zhang, Qin Lu, Liang Qiu, Changlong Yu, Puyang Xu, Chao Zhang, Bing Yin, Hyokun Yun, Lihong Li

Challenge:	Existing work on reinforcement learning has focused on single-turn tasks such as solving math problems.
Approach:	They propose a framework that learns directly from online interactions by asynchronously generating diverse trajectories, guided by binary rewards depending on task success.
Outcome:	Experiments on the WebArena-Lite benchmark show that the framework outperforms state-of-the-art methods and strong proprietary models.

ConCISE: Confidence-guided Compression in Step-by-step Efficient Reasoning (2025.emnlp-main)

Copied to clipboard

Ziqing Qiao, Yongheng Deng, Jiali Zeng, Dong Wang, Lai Wei, Guanbo Wang, Fandong Meng, Jie Zhou, Ju Ren, Yaoxue Zhang

Unveiling Factual Recall Behaviors of Large Language Models through Knowledge Neurons (2024.emnlp-main)

Challenge:	Existing methods for fine-tuning-based compression suffer from verbose outputs, increasing computational overhead.
Approach:	They propose a framework to generate concise reasoning chains using Confidence Injection and Early Stopping.
Outcome:	The proposed framework reduces the length of the model by up to 50% while maintaining high task accuracy.

Copied to clipboard

Yifei Wang, Yuheng Chen, Wanting Wen, Yu Sheng, Linjing Li, Daniel Zeng

UBench: Benchmarking Uncertainty in Large Language Models with Multiple Choice Questions (2025.findings-acl)

Challenge:	Recent advances in Large Language Models have underscored their exceptional reasoning prowess with natural language understanding across a broad spectrum of tasks.
Approach:	They examine whether Large Language Models actively recall or retrieve their internal repositories of factual knowledge when faced with reasoning tasks.
Outcome:	The proposed model improves reasoning performance while suppressing it leads to notable degradation.

Copied to clipboard

Xunzhi Wang, Zhuowei Zhang, Gaonan Chen, Qiongyu Li, Bitong Luo, Zhixin Han, Haotian Wang, Zhiyu Li, Hang Gao, Mengting Hu

LLMEmbed: Rethinking Lightweight LLM’s Genuine Function in Text Classification (2024.acl-long)

Challenge:	Existing methods for benchmarking the uncertainty of large language models face challenges . existing methods require internal model access, additional training, or high computational costs .
Approach:	They propose a new benchmark for evaluating the uncertainty of large language models based on confidence intervals . UBench encompasses 11,978 multiple choice questions spanning knowledge, language, understanding, and reasoning capabilities.
Outcome:	The proposed method outperforms existing methods for benchmarking the uncertainty of large language models.

Copied to clipboard

ChunLiu ChunLiu, Hongguang Zhang, Kainan Zhao, Xinghai Ju, Lin Yang

BBA: Bi-Modal Behavioral Alignment for Reasoning with Large Vision-Language Models (2024.findings-acl)

Challenge:	Recent attempts to improve text classification performance are based on heuristic Chain-of-Thought (CoT) LLMEmbed is a simple and effective transfer learning strategy that can be used to improve the performance of large language models.
Approach:	They propose a simple transfer learning strategy to improve text classification using heuristic Chain-of-Thought.
Outcome:	The proposed method achieves strong performance on publicly available datasets while using low training overhead.

Copied to clipboard

Xueliang Zhao, Xinting Huang, Tingchen Fu, Qintong Li, Shansan Gong, Lemao Liu, Wei Bi, Lingpeng Kong

Distilling Reasoning Capabilities into Smaller Language Models (2023.findings-acl)

Challenge:	Multimodal reasoning is a key capability for large vision-language models . however, the vanilla Chain-of-Thought method fails to address critical steps in multi-step reasoning tasks.
Approach:	They propose a bi-modal Behavioral Alignment method to augment multimodal reasoning . they use domain-specific language to integrate multimodal information into a precise alternative form .
Outcome:	The proposed method significantly improves GPT-4V(ision) on geometry problem solving, chess positional advantage prediction and molecular property prediction.

Copied to clipboard

Kumar Shridhar, Alessandro Stolfo, Mrinmaya Sachan

Retrieval and Reasoning on KGs: Integrate Knowledge Graphs into Large Language Models for Complex Question Answering (2024.findings-emnlp)

Challenge:	a step-by-step reasoning approach like chain of thought has proved to be effective in eliciting reasoning abilities in large language models.
Approach:	They propose a knowledge distillation approach that leverages CoT reasoning capabilities of larger models and distills them into smaller models.
Outcome:	The proposed scheme boosts the performance of smaller models over 70% on multiple reasoning datasets.

Copied to clipboard

Yixin Ji, Kaixin Wu, Juntao Li, Wei Chen, Mingjie Zhong, Xu Jia, Min Zhang

Vision-Language Models Can Self-Improve Reasoning via Reflection (2025.naacl-long)

Challenge:	Large Language Models (LLMs) have performed impressively in various NLP tasks, but their inherent hallucination phenomena severely challenge their credibility in complex reasoning.
Approach:	They propose to integrate explainable Knowledge Graphs (KGs) with LLMs to alleviate hallucinations . they construct subgraphs to enhance the retrieval capabilities of KGs via CoT reasoning.
Outcome:	Extensive experiments on two KGQA datasets show that the proposed model achieves convincing performance compared to strong baselines.

Copied to clipboard

Kanzhi Cheng, Li YanTao, Fangzhi Xu, Jianbing Zhang, Hao Zhou, Yang Liu

Mitigating Demonstration Bias through Global Coevolutionary Reasoning (2025.findings-acl)

Challenge:	Chain-of-thought (CoT) has been shown to improve the reasoning capability of large language models (LLMs).
Approach:	They propose a framework which iteratively enhances the model’s Vision-language Reasoning by Reflecting on CoT Rationales.
Outcome:	The proposed framework improves multimodal reasoning on vision-language tasks by 23% to 60% over baselines.

Copied to clipboard

Chuan Gou, Bangwei Li, Jianhua Dai, Xiaoyang Han, Ming Cai

Think Twice: Perspective-Taking Improves Large Language Models’ Theory-of-Mind Capabilities (2024.acl-long)

Challenge:	Existing methods for chain-of-thought prompting rely on manual demonstrations . experimental results show that GCR outperforms baseline methods without performance degradation .
Approach:	They propose a method that uses random samples to generate demonstrations in zero-shot settings.
Outcome:	The proposed method outperforms baseline methods on ten datasets without demonstration bias.

Copied to clipboard

Alex Wilf, Sihyun Lee, Paul Pu Liang, Louis-Philippe Morency

FedCoT: Federated Chain-of-Thought Distillation for Large Language Models (2025.findings-emnlp)

Challenge:	Recent advances to LLMs’ reasoning capabilities from simple yet effective prompting techniques such as Chain-of-Thought have seen limited applicability to ToM.
Approach:	They propose a two-stage prompting framework inspired by Simulation Theory's notion of perspective-taking to elicit Theory-of-Mind capabilities in Large Language Models.
Outcome:	The proposed framework shows that it is much more effective than existing prompts.

Copied to clipboard

Tao Fan, Weijing Chen, Yan Kang, Guoqiang Ma, Hanlin Gu, Yuanfeng Song, Lixin Fan, Qiang Yang

SalaMAnder: Shapley-based Mathematical Expression Attribution and Metric for Chain-of-Thought Reasoning (2025.findings-emnlp)

Challenge:	Large Language Models (LLMs) have emerged as a transformative force in artificial intelligence, demonstrating exceptional proficiency across various tasks.
Approach:	They propose a federated framework for the Chain-of-Thought distillation of knowledge from LLMs to SLMs, while adhering to privacy requirements.
Outcome:	The proposed framework ensures secure knowledge transfer from an LLM on a high-powered server to an SLM on resource-constrained client while adhering to privacy requirements.

Copied to clipboard

Yue Xin, Chen Shen, Shaotian Yan, Xiaosong Yuan, Yaoming Wang, Xiaofeng Zhang, Chenxi Huang, Jieping Ye

Revisiting Large Language Models as Zero-shot Relation Extractors (2023.findings-emnlp)

Challenge:	Chain-of-Thought prompting improves the math reasoning capability of large language models.
Approach:	They propose a method for attribution of component-level contributions in CoT reasoning using Shapley value and a stratified sampling algorithm that significantly reduces computational complexity.
Outcome:	The proposed method reduces computational complexity and provides robust correlations with model performance.

Copied to clipboard

Guozheng Li, Peng Wang, Wenjun Ke

FastCuRL: Curriculum Reinforcement Learning with Stage-wise Context Scaling for Efficient Training R1-like Reasoning Models (2025.findings-emnlp)

Challenge:	Recent studies show that large language models (LLMs) transfer well to new tasks out-of-the-box . relationship extraction (RE) involves a certain degree of labeled or unlabeled data even under zero-shot setting.
Approach:	They propose a simple prompt recursively using LLMs to transform RE inputs to QA format . they propose qq prompting and qt prompting to improve their results .
Outcome:	The proposed method improves on different model sizes, benchmarks and settings.

Copied to clipboard

Mingyang Song, Mao Zheng, Zheng Li, Wenjie Yang, Xuan Luo

TCPO: Thought-Centric Preference Optimization for Effective Embodied Decision-making (2025.emnlp-main)

Challenge:	Improving training efficiency remains a challenge in large-scale Reinforcement Learning (RL).
Approach:	They propose a curriculum RL framework with stage-wise context scaling to improve RL training efficiency.
Outcome:	The proposed framework outperforms state-of-the-art reasoning models on five benchmarks and achieves 49.6% accuracy on AIME 2024.

Copied to clipboard

Kechen Jiao, Zhirui Fang, Jiahao Liu, Bei Li, Qifan Wang, Xinyu Liu, Junhao Ruan, Zhongjian Qiao, Yifan Zhu, Yaxin Xu, Jingang Wang, Xiu Li

Critical-CoT: A Robust Defense Framework against Reasoning-Level Backdoor Attacks in Large Language Models (2026.acl-long)

Challenge:	Existing post-SFT methods for embodied AI are constrained by sparse rewards and action-only optimization, resulting in low sample efficiency, poor consistency, and model degradation.
Approach:	They propose to integrate Thought-Centric Preference Optimization (TCPO) into embodied decision-making by transforming sparse reward signals into richer step sample pairs.
Outcome:	The proposed approach achieves an average success rate of 26.67% in the ALFWorld environment, and a 6% improvement over RL4VLM.

Copied to clipboard

Vu Tuan Truong, Long Bao Le

Focus on Your Question! Interpreting and Mitigating Toxic CoT Problems in Commonsense Reasoning (2024.acl-long)

Challenge:	Large Language Models (LLMs) are vulnerable to backdoors that use long-form reasoning to generate a specific word, choice, or class.
Approach:	They propose a mechanism that allows LLMs to develop critical thinking behaviors and detect backdoors by a two-stage fine-tuning.
Outcome:	The proposed mechanism exhibits strong cross-domain and cross-task generalization.

Copied to clipboard

Jiachun Li, Pengfei Cao, Chenhao Wang, Zhuoran Jin, Yubo Chen, Daojian Zeng, Kang Liu, Jun Zhao

DialCoT Meets PPO: Decomposing and Exploring Reasoning Paths in Smaller Language Models (2023.emnlp-main)

Challenge:	Large language models exhibit high-level commonsense reasoning abilities, especially with enhancement methods like Chain-of-Thought (CoT).
Approach:	They propose a chain-of-thought-like method to elicit models' potential abilities to generate rationales and answers that are based on attribution tracing and causal tracers to probe the internal working mechanism of the LLM.
Outcome:	The proposed method eliminates Toxic CoT problems and improves the model’s overall commonsense reasoning performance by 5.5%.

Copied to clipboard

Chengcheng Han, Xiaowei Du, Che Zhang, Yixin Lian, Xiang Li, Ming Gao, Baoyuan Wang

Dialogue Systems & Conversational Agents

Challenge:	Chain-of-Thought prompting has improved the reasoning capabilities of Large Language Models (LLMs) but it is ineffective or detrimental to the performance on reasoning tasks in Smaller Language Model (SLMs) with less than 10 billion parameters.
Approach:	They propose a Dialogue-guided Chain-of-Thought method to improve the reasoning capabilities of Large Language Models (LLMs) by generating intermediate reasoning steps in a dialogue format to guide the model to the final answer.
Outcome:	The proposed method can achieve significant performance gains over state-of-the-art competitors on four arithmetic reasoning datasets.

Measuring Chain of Thought Faithfulness by Unlearning Reasoning Steps (2025.emnlp-main)

Copied to clipboard

Martin Tutek, Fateme Hashemi Chaleshtori, Ana Marasovic, Yonatan Belinkov

Reasoning with Language Model is Planning with World Model (2023.emnlp-main)

Challenge:	Language models (LMs) produce a chain of thought (CoT) when prompted to think step-by-step, but it is unclear whether the reasoning encoded in the CoT is faithful to the models’ parametric beliefs.
Approach:	They propose a framework for measuring parametric faithfulness of generated reasoning by unlearning reasoning steps (FUR) they propose to erase information contained in reasoning steps from model parameters and measure faithfulness as the resulting effect on the model’s prediction.
Outcome:	The proposed framework erases information contained in reasoning steps from model parameters and measures faithfulness as the resulting effect on the model’s prediction.

Copied to clipboard

Shibo Hao, Yi Gu, Haodi Ma, Joshua Hong, Zhen Wang, Daisy Wang, Zhiting Hu

“Well, Keep Thinking”: Enhancing LLM Reasoning with Adaptive Injection Decoding (2025.findings-acl)

Challenge:	Large language models (LLMs) have shown remarkable reasoning capabilities, particularly with Chain-of-Thought-style prompts.
Approach:	They propose a framework that repurposes the LLM as both a world model and a reasoning agent and incorporates a principled planning algorithm (based on Monte Carlo Tree Search)
Outcome:	The proposed framework repurposes the LLM as both a world model and a reasoning agent and incorporates a principled planning algorithm (based on Monte Carlo Tree Search) it achieves optimum balance between exploration and exploitation, while achieving high-reward reasoning paths efficiently.

Copied to clipboard

Hyunbin Jin, Je Won Yeom, Seunghyun Bae, Taesup Kim

Revisiting Parallel Context Windows: A Frustratingly Simple Alternative and Chain-of-Thought Deterioration (2024.findings-acl)

Challenge:	Large language models (LLMs) exhibit strong reasoning abilities, often attributed to few-shot or zero-shot Chain-of-Thought (CoT) prompting.
Approach:	They propose a decoding strategy that nudges LLMs to continue reasoning, thereby preventing immature reasoning processes.
Outcome:	The proposed method significantly improves LLM reasoning capabilities on diverse reasoning benchmarks.

Copied to clipboard

Kejuan Yang, Xiao Liu, Kaiwen Men, Aohan Zeng, Yuxiao Dong, Jie Tang

Feature Extraction and Steering for Enhanced Chain-of-Thought Reasoning in Language Models (2025.emnlp-main)

Challenge:	Existing methods for extending the maximum context lengths of language models are lacking a strong baseline for in-context few-shot classification and on more challenging Chain-of-Thought reasoning, such as HotpotQA, deteriorate question miscomprehension and false inference.
Approach:	They propose to harness window-wise attention and positional embedding techniques to extend the maximum context lengths of language models.
Outcome:	The proposed method is able to extend the maximum context lengths of language models, e.g., 2048 for LLaMA, by harnessing window-wise attention and positional embedding techniques.

Copied to clipboard

Zihao Li, Xu Wang, Yuzhe Yang, Ziyu Yao, Haoyi Xiong, Mengnan Du

SoRFT: Issue Resolving with Subtask-oriented Reinforced Fine-Tuning (2025.acl-long)

Challenge:	Large Language Models (LLMs) can solve reasoning and mathematical problems using the Chain-of-Thought technique, but require costly and long CoT data and fine-tuning.
Approach:	They propose a method that uses Sparse Autoencoders to extract interpretable features from vanilla CoT and use them to steer the LLM's internal states.
Outcome:	The proposed method uses Sparse Autoencoders (SAEs) to extract interpretable features from vanilla CoT and steer the LLM's internal states during generation.

Copied to clipboard

Zexiong Ma, Chao Peng, Pengfei Gao, Xiangxin Meng, Yanzhen Zou, Bing Xie

CommonIT: Commonality-Aware Instruction Tuning for Large Language Models via Data Partitions (2024.emnlp-main)

Challenge:	Existing issue-resolving frameworks rely on commercial models, leading to high costs and privacy concerns.
Approach:	They propose a training approach to enhance issue resolving capability of LLMs by decomposing issue reasolving into subtasks.
Outcome:	The proposed approach improves issue-resolving performance and generalizes model . it is cost-effective and provides a cost-efficient alternative to commercial models .

Copied to clipboard

Jun Rao, Xuebo Liu, Lian Lian, Shengjun Cheng, Yunjie Liao, Min Zhang

Large Language Models Can Learn Temporal Reasoning (2024.acl-long)

Challenge:	Current studies have focused on fine-tuning, but the use of instruction tuning is not as effective as fine-cuning.
Approach:	They propose a commonality-aware instruction tuning strategy to cluster instruction datasets into distinct groups with three proposed metrics Task, Embedding and Length.
Outcome:	The proposed strategy boosts an average improvement of 2.1% on the general domain and 5.2% on the special domain.

Copied to clipboard

Siheng Xiong, Ali Payani, Ramana Kompella, Faramarz Fekri

AgentThink: A Unified Framework for Tool-Augmented Chain-of-Thought Reasoning in Vision-Language Models for Autonomous Driving (2025.findings-emnlp)

Challenge:	Temporal reasoning (TR) is a fundamental ability of large language models (LLMs) however, there is neo-standard methods to perform TR, which are not suitable for large language model applications.
Approach:	They propose a framework to enhance temporal reasoning by using a latent representation, temporal graph (TG) instead of reasoning over the original context, they adopt a temporal representation that enhances TR learning.
Outcome:	The proposed framework improves the learning of language-based TR by incorporating a latent representation, temporal graph (TG) a synthetic dataset is constructed for fine-tuning LLMs on text-to-TG translation tasks and benchmarks.

Copied to clipboard

Kangan Qian, Sicong Jiang, Yang Zhong, Ziang Luo, Zilin Huang, Tianze Zhu, Kun Jiang, Mengmeng Yang, Zheng Fu, Jinyu Miao, Yining Shi, He Zhe Lim, Li Liu, Tianbao Zhou, Hongyi Wang, Huang Yu, Yifei Hu, Guang Li, Guang Chen, Hao Ye, Lijun Sun, Diange Yang

CLAMBER: A Benchmark of Identifying and Clarifying Ambiguous Information Needs in Large Language Models (2024.acl-long)

Challenge:	Vision-Language Models struggle with hallucinations, inefficient reasoning, and limited real-world validation hinders accurate perception and robust step-by-step reasoning.
Approach:	AgentThink integrates Chain-of-Thought reasoning with dynamic, agent-style tool invocation for autonomous driving tasks.
Outcome:	Experiments on the DriveLMM-o1 benchmark show AgentThink significantly boosts overall reasoning scores by 53.91% and enhances answer accuracy by 33.54% .

Copied to clipboard

Tong Zhang, Peixin Qin, Yang Deng, Chen Huang, Wenqiang Lei, Junhong Liu, Dingnan Jin, Hongru Liang, Tat-Seng Chua

GODBench: A Benchmark for Multimodal Large Language Models in Video Comment Art (2025.acl-long)

Challenge:	Large language models are used to meet user information needs, but their effectiveness in dealing with user queries that contain various types of ambiguity remains unknown.
Approach:	They propose a benchmark for evaluating large language models using a well-organized taxonomy.
Outcome:	The proposed model is based on a well-organized taxonomy and compares it with other models.

Copied to clipboard

Yiming Lei, Chenkai Zhang, Zeming Liu, Haitao Leng, ShaoGuo Liu, Tingting Gao, Qingjie Liu, Yunhong Wang

Reasoning Robustness of LLMs to Adversarial Typographical Errors (2024.emnlp-main)

Challenge:	Existing benchmarks for video comment art are constrained by their limited modalities and insufficient categories, hindering creativity in video-based comment art creation.
Approach:	They propose a benchmark that integrates video and text modalities to evaluate MLLMs’ abilities to compose video Comment art.
Outcome:	The proposed framework integrates video and text modalities to evaluate MLLMs’ abilities to compose video comment art.

Copied to clipboard

Esther Gan, Yiran Zhao, Liying Cheng, Mao Yancan, Anirudh Goyal, Kenji Kawaguchi, Min-Yen Kan, Michael Shieh

Robustness in NLP

Challenge:	Large Language Models (LLMs) have demonstrated impressive capabilities in reasoning using Chain-of-Thought (CoT) prompting.
Approach:	They develop an algorithm that iteratively samples typos for words that are important to the query and selects the edit that is most likely to succeed in attacking.
Outcome:	The proposed algorithm detects typographical errors in large and closed-source LLMs and shows that they are robust to them.

Distilling the Essence: Efficient Reasoning Distillation via Sequence Truncation (2026.findings-acl)

Copied to clipboard

Wei-Rui Chen, Vignesh Kothapalli, Ata Fatahibaarzi, Hejian Sang, Shao Tang, Qingquan Song, Zhipeng Wang, Muhammad Abdul-Mageed

Safe: Enhancing Mathematical Reasoning in Large Language Models via Retrospective Step-aware Formal Verification (2025.acl-long)

Challenge:	a large reasoning model (LRM) training on large amounts of reasoning data is computationally expensive.
Approach:	They propose a method to quantify computation-quality tradeoffs as a function of sequence length.
Outcome:	The proposed method reduces training time, memory and FLOPs by 50% on long training sequences while retaining the full-sequence performance.

Copied to clipboard

Chengwu Liu, Ye Yuan, Yichun Yin, Yan Xu, Xin Xu, Zaoyu Chen, Yasheng Wang, Lifeng Shang, Qun Liu, Ming Zhang

Automatic Mathematic In-Context Example Generation for LLM Using Multi-Modal Consistency (2025.coling-main)

Challenge:	Chain-of-Thought prompting is a de facto method to elicit reasoning capabilities from large language models (LLMs).
Approach:	They propose a step-aware formal verification framework Safe to address hallucinations in CoT prompting . they propose 'formal step' as a benchmark for step correctness theorem proving with 30,809 formal statements.
Outcome:	The proposed framework shows significant performance improvement while offering interpretable and verifiable evidence.

Copied to clipboard

Jaeseong Lee, Wei Yang, Gopal Gupta, Shiyi Wei

DS-MHP: Improving Chain-of-Thought through Dynamic Subgraph-Guided Multi-Hop Path (2025.findings-emnlp)

Challenge:	Existing methods for in-context learning require annotated datasets, resulting in higher computational costs and lower quality examples.
Approach:	They propose a framework that automatically generates high-quality in-context examples to enhance LLMs’ mathematical reasoning.
Outcome:	Evaluated on four math problem datasets, the proposed framework outperforms baseline methods with LLM accuracy ranging from 87.0% to 99.3%.

Copied to clipboard

Yongqiang Liu, Qiyao Peng, Binrong Liu, Hongtao Liu, XueWei Li, Wenjun Wang

Retrieving, Rethinking and Revising: The Chain-of-Verification Can Improve Retrieval Augmented Generation (2024.findings-emnlp)

Challenge:	Existing knowledge graph methods lack adaptability in knowledge-intensive tasks with multiple entities and implicit multi-hop relations.
Approach:	They propose a zero-shot framework to enhance LLM reasoning in multi-entity relation tasks.
Outcome:	DS-MHP outperforms baselines and state-of-the-art methods on 12 datasets spanning commonsense, logical, symbolic, and arithmetic reasoning.

Copied to clipboard

Bolei He, Nuo Chen, Xinran He, Lingyong Yan, Zhenkai Wei, Jinchang Luo, Zhen-Hua Ling

Enhancing the Reasoning Capabilities of Small Language Models via Solution Guidance Fine-Tuning (2025.coling-main)

Challenge:	Recent Retrieval Augmented Generation (RAG) aims to enhance Large Language Models . however, such approach can generate inconsistent answer with external references .
Approach:	They propose to integrate the verification module into the RAG to improve external retrieval correctness and internal generation consistency.
Outcome:	The proposed model can significantly surpass the state-of-the-art baselines using different LLM backbones.

Copied to clipboard

Jing Bi, Yuting Wu, Weiwei Xing, Zhenjie Wei

Facilitating Long Context Understanding via Supervised Chain-of-Thought Reasoning (2025.emnlp-main)

Challenge:	Large language models (LLMs) have demonstrated remarkable performance across a wide range of tasks.
Approach:	They propose a new reasoning strategy Solution Guidance (SG) and a plug-and-play training paradigm Solution-Guidance Fine-Tuning (SGFT) which focuses on problem understanding and decomposition at the semantic and logical levels, rather than specific computations.
Outcome:	The proposed reasoning strategy Solution Guidance (SG) and plug-and-play training paradigm Solution-Guidance Fine-Tuning (SGFT) improves the reasoning capabilities of small language models on various reasoning tasks.

Copied to clipboard

Jingyang Lin, Andy Wong, Tian Xia, Shenghua He, Hui Wei, Mei Han, Jiebo Luo

Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning (2026.findings-acl)

Challenge:	Recent advances in Large Language Models (LLMs) have enabled them to process increasingly longer sequences, ranging from 2K to 2M tokens and even beyond.
Approach:	They propose a synthetic dataset in the financial domain that integrates Chain-of-Thought reasoning into LLMs in a supervised manner to facilitate effective long-context understanding.
Outcome:	The proposed model outperforms standard GPT-4o-mini on the Loong benchmark and fine tunes LLaMA-3.1-8B-Instruct on the model, achieving a 28.0% gain on the financial subset.

Copied to clipboard

Nathanaël Carraz Rakotonirina, Ren Pang, Neha Anna John, Michael Bohlke-Schneider, Momchil Hardalov

AS-ES Learning: Towards efficient CoT learning in small models (2024.findings-acl)

Challenge:	Large language models (LLMs) increase test-time computation, often in the form of chain-of-thought (CoT) however, reasoning traces can become unnecessarily long, increasing computation costs without improving accuracy and sometimes even degrading performance.
Approach:	They propose a multi-stage efficient reasoning method that combines supervised fine-tuning with reinforcement learning using an adaptive length penalty.
Outcome:	The proposed method reduces response length by an average of 28% for 8B models and 40% for 32B models while incurring only minor performance drops of 1.6 and 2.5 points, respectively.

Copied to clipboard

Nuwa Xi, Yuhan Chen, Sendong Zhao, Haochun Wang, GongZhang GongZhang, Bing Qin, Ting Liu

CoT-VTM: Visual-to-Music Generation with Chain-of-Thought Reasoning (2025.findings-acl)

Challenge:	Existing methods to induce Chain-of-Thought (CoT) in LLMs are limited and do not consider the importance of efficiently utilizing existing CoT data.
Approach:	They propose a new training paradigm which exploits the inherent information in CoT for iterative generation.
Outcome:	The proposed training paradigm surpasses direct seq2seq training on CoT-extensive tasks without data augmentation or altering the model itself.

Copied to clipboard

Xikang Guan, Zheng Gu, Jing Huo, Tianyu Ding, Yang Gao

mCoT: Multilingual Instruction Tuning for Reasoning Consistency in Language Models (2024.acl-long)

Challenge:	Existing methods for visual-to-music generation lack large-scale, high-quality visual-music paired datasets and lack of direct semantic correspondence between visuals and music.
Approach:	They propose a framework that distills Chain-of-Thought reasoning to enable visual-to-music generation without paired data.
Outcome:	The proposed framework achieves optimal performance on image-to-music and video-to music tasks.

Copied to clipboard

Huiyuan Lai, Malvina Nissim

Knowing Before Saying: LLM Representations Encode Information About Chain-of-Thought Success Before Completion (2025.findings-acl)

Challenge:	Existing models show low performance for lesser resourced languages, but they can achieve surprising performance on complex reasoning tasks in natural language processing (NLP).
Approach:	They compile the first large-scale multilingual math reasoning dataset, mCoT-MATH, covering eleven diverse languages.
Outcome:	The proposed model achieves impressive consistency across languages and comparable performance to close- and open-source models even of much larger sizes.

Copied to clipboard

Anum Afzal, Florian Matthes, Gal Chechik, Yftah Ziser

Finite State Automata Inside Transformers with Chain-of-Thought: A Mechanistic Study on State Tracking (2025.acl-long)

Challenge:	Using later reasoning steps does not always improve classification, suggesting LLMs encode key information early.
Approach:	They propose a method to predict the success of a zero-shot Chain-of-Thought process by using LLM representations that are based on initial steps representations.
Outcome:	The proposed method performs well even before a single token is generated, suggesting that crucial information about the reasoning process is already present in the initial steps representations.

Copied to clipboard

Yifan Zhang, Wenyu Du, Dongming Jin, Jie Fu, Zhi Jin

Leveraging Training Data in Few-Shot Prompting for Numerical Reasoning (2023.findings-acl)

Challenge:	Existing studies show that Chain-of-thought (CoT) can enhance the performance of large language models (LLMs) however, there is limited understanding of the algorithms that Transformer+CoT can learn.
Approach:	They propose two metrics to evaluate Transformer+CoT's state tracking capabilities and identify the circuit responsible for tracking the world state.
Outcome:	The proposed model achieves 100% accuracy for each state, highlighting an implicit finite state automaton (FSA) embedded within the model.

Copied to clipboard

Zhanming Jie, Wei Lu

MAmmoTH-VL: Eliciting Multimodal Reasoning with Instruction Tuning at Scale (2025.acl-long)

Challenge:	Chain-of-thought (CoT) prompts can be challenging to design for arithmetic word problem solving.
Approach:	They propose to use training data to replace CoT with programs as the reasoning step . their results show that leveraging training data can improve generalization ability of prompts .
Outcome:	The proposed methods improve the generalization ability of prompts and the performance of fine-tuned smaller models in arithmetic word problem solving.

Copied to clipboard

Jiawei Guo, Tianyu Zheng, Yizhi Li, Yuelin Bai, Bo Li, Yubo Wang, King Zhu, Graham Neubig, Wenhu Chen, Xiang Yue

MuTIS: Enhancing Reasoning Efficiency through Multi Turn Intervention Sampling in Reinforcement Learning (2025.emnlp-main)

Challenge:	Current instruction-tuning datasets focus on simplistic visual question answering tasks, and provide phrase-level answers without any intermediate rationales.
Approach:	They propose to use open-source multimodal large language models to train MLLMs on a dataset with 12M instruction-response pairs to elicit CoT reasoning.
Outcome:	The proposed model achieves state-of-the-art performance on benchmarks such as MathVerse, MMMU-Pro, and MuirBench, and gains improvements of up to 4% on non-reasoning-based benchmarks.

Copied to clipboard

Wenshuo Zhao, Haoxing Zhai, Xinyu Qiu, Zhenting Qi, Shuhe Li, Linchao Zhu

KnowledgeFMath: A Knowledge-Intensive Math Reasoning Dataset in Finance Domains (2024.acl-long)

Challenge:	Existing methods for training large reasoning models with long chain-of-thought (CoT) are limited by the number of parameters and the complexity of the model.
Approach:	They propose a framework that leverages multi-turn interventions to produce concise reasoning chains and demonstrates strong scalability.
Outcome:	The proposed framework breaks the accuracy-efficiency trade-off by producing concise reasoning chains and demonstrating strong scalability on 7B models.

Copied to clipboard

Yilun Zhao, Hongjun Liu, Yitao Long, Rui Zhang, Chen Zhao, Arman Cohan

ChartM3: A Multi-Stage Code-Driven Pipeline for Constructing Multi-Dimensional and Multi-Step Visual Reasoning Data in Chart Comprehension (2025.findings-emnlp)

Challenge:	Existing benchmarks for large language models (LLMs) are only 56.6% accurate, leaving room for improvement.
Approach:	They propose a benchmark to evaluate LLMs' capabilities in solving knowledge-intensive math reasoning problems using a finance-domain knowledge bank and expert-annotated solution references.
Outcome:	The proposed system achieves only 56.6% accuracy, leaving room for improvement.

Copied to clipboard

Duo Xu, Hao Cheng, Xin Lin, Zhen Xie, Hao Henry Wang

Beyond Chain-of-Thought: A Survey of Chain-of-X Paradigms for LLMs (2025.coling-main)

Challenge:	Currently, research on complex chart understanding tasks is limited . a pipeline for visual reasoning datasets addresses these limitations .
Approach:	They propose a code-driven pipeline for generating visual reasoning datasets . pipeline integrates retrieval-augmented generation to retrieve professional chart templates .
Outcome:	The proposed pipeline enhances chart diversity and data quality through model-based evaluation.

Copied to clipboard

Yu Xia, Rui Wang, Xu Liu, Mingyan Li, Tong Yu, Xiang Chen, Julian McAuley, Shuai Li

Challenge:	Large Language Models (LLMs) have shown impressive reasoning abilities when prompted with Chain-of-Thought (CoT).
Approach:	They propose to categorize Chain-of-X methods by taxonomies of nodes, i.e., the X in CoX, and application tasks, and then categorise them by taxanomies and discuss potential future directions.
Outcome:	The proposed methods are categorised by taxonomies of nodes, i.e., the X in CoX, and application tasks.

A Closer Look at Bias and Chain-of-Thought Faithfulness of Large (Vision) Language Models (2025.findings-emnlp)

Copied to clipboard

Sriram Balasubramanian, Samyadeep Basu, Soheil Feizi

Can LLMs Reason Abstractly Over Math Word Problems Without CoT? Disentangling Abstract Formulation From Arithmetic Computation (2025.emnlp-main)

Challenge:	Chain-of-thought reasoning improves performance of large language models, but is it faithfully reflecting internal processes?
Approach:	They propose a new evaluation pipeline for categorizing bias articulation patterns and a novel evaluation pipeline to examine CoT faithfulness in large vision-language models.
Outcome:	The proposed evaluation pipeline enables significantly more precise analysis of CoT reasoning than previous methods.

Copied to clipboard

Ziling Cheng, Meng Cao, Leila Pishdad, Yanshuai Cao, Jackie CK Cheung

Not All Tokens Matter: Towards Efficient LLM Reasoning via Token Significance in Reinforcement Learning (2026.acl-long)

Challenge:	Large language models (LLMs) are often evaluated on math word problems . however, such metrics conflate two distinct sub-skills: abstract formulation and arithmetic computation.
Approach:	They propose to use Final-answer-based metrics to evaluate large language models on math word problems to conflate two distinct sub-skills: abstract formulation and arithmetic computation.
Outcome:	The proposed model performance is bottlenecked by arithmetic computation and not abstract formulation, the study shows.

Copied to clipboard

Hanbing Liu, Lang Cao, Yuanyi Ren, Mengyu Zhou, Haoyu Dong, Xiaojun Ma, Shi Han, Dongmei Zhang

Revisiting Chain-of-Thought Prompting: Zero-shot Can Be Stronger than Few-shot (2025.findings-emnlp)

Challenge:	Large language models (LLMs) often produce unnecessarily long explanations that reduce efficiency.
Approach:	They propose a length-aware reward that selectively penalizes insignificance tokens . they also propose 'dynamic length control' that encourages more detailed reasoning .
Outcome:	The proposed method reduces response length while maintaining correctness, the authors show . it selectively penalizes insignificance tokens while maintaining accuracy .

Copied to clipboard

Xiang Cheng, Chengyan Pan, Minjun Zhao, Deyang Li, Fangchao Liu, Xinyu Zhang, Xiao Zhang, Yong Liu

MMMU-Pro: A More Robust Multi-discipline Multimodal Understanding Benchmark (2025.acl-long)

Challenge:	In-Context Learning (ICL) is an essential emergent ability of Large Language Models (LLMs).
Approach:	They introduce CoT to exemplars of ICL to enhance the reasoning capability . however, it remains unclear whether CoT exemplar is still beneficial for recent, stronger models in such tasks.
Outcome:	The enhanced exemplars fail to improve the model’s reasoning performance, despite being constructed using answers from advanced models such as Qwen2.5-Max and DeepSeek-R1.

Copied to clipboard

Xiang Yue, Tianyu Zheng, Yuansheng Ni, Yubo Wang, Kai Zhang, Shengbang Tong, Yuxuan Sun, Botao Yu, Ge Zhang, Huan Sun, Yu Su, Wenhu Chen, Graham Neubig

LoRA-PAR: A Flexible Dual-System LoRA Partitioning Approach to Efficient LLM Fine-Tuning (2025.findings-emnlp)

Challenge:	Recent advances in multimodal large language models have led to progress in tackling complex reasoning tasks that combine textual and visual information.
Approach:	They introduce a robust version of the Massive Multi-discipline Multimodal Understanding and Reasoning (MMMU) benchmark.
Outcome:	The proposed model performs lower on MMMU-Pro than on the previous benchmark, ranging from 16.8% to 26.9%.

Copied to clipboard

Yining Huang, Bin Li, Keke Tang, Meilian Chen

Guided by Gut: Efficient Test-Time Scaling with Reinforced Intrinsic Confidence (2026.acl-long)

Challenge:	Large-scale generative models like DeepSeek-R1 and OpenAI-O1 benefit substantially from chain-of-thought reasoning, yet pushing their performance typically requires vast data, large model sizes, and full-parameter fine-tuning.
Approach:	They propose a dual-system LoRA framework that partitions data and parameters by System 1 or System 2 demands and adopts a two-stage fine-tuning strategy to enhance knowledge and intuition.
Outcome:	The proposed framework partitions data and parameters by System 1 or System 2 demands, using fewer yet more focused parameters for each task.

Copied to clipboard

Amirhosein Ghasemabadi, Keith G. Mills, Baochun Li, Di Niu

How Many Demonstrations Do You Need for In-context Learning? (2023.findings-emnlp)

Challenge:	Guided by Gut (GG) is an efficient self-guided TTS framework for Large Language Models (LLMs) that performs step-by-step reasoning at a low cost without any reward models or verifiers.
Approach:	They propose a self-guided TTS framework that enables LLMs to perform step-by-step reasoning at a low cost without any reward models or verifiers.
Outcome:	Empirical evaluations show that GG performs better than TTS with PRMs while reducing GPU memory usage by up to 10.

Copied to clipboard

Jiuhai Chen, Lichang Chen, Chen Zhu, Tianyi Zhou

Is Chain-of-Thought Reasoning of LLMs a Mirage? A Data Distribution Lens (2026.findings-acl)

Challenge:	Large language models (LLMs) are capable of complex reasoning when given a few input-output demos.
Approach:	They use fewer input-output demos for each test query to study ICL . they do not observe significant degradation when using only one randomly chosen demo .
Outcome:	The proposed model outperforms multi-demo models on the tasks in 2022.

Copied to clipboard

Chengshuai Zhao, Zhen Tan, Pingchuan Ma, Dawei Li, Bohan Jiang, Yancheng Wang, Yingzhen Yang, Huan Liu

Fewer is More: Boosting Math Reasoning with Reinforced Context Pruning (2024.emnlp-main)

Challenge:	Chain-of-Thought (CoT) prompting has been shown to be effective in eliciting structured reasoning from large language models (LLMs).
Approach:	They propose a data distribution lens to understand when and why CoT reasoning fails . they propose 'data-based' training that trains LLMs from scratch .
Outcome:	The proposed model enables models to generate reasoning trajectories that approximate those observed during training.

Copied to clipboard

Xijie Huang, Li Lyna Zhang, Kwang-Ting Cheng, Fan Yang, Mao Yang

Self-Polish: Enhance Reasoning in Large Language Models via Problem Refinement (2023.findings-emnlp)

Challenge:	Large Language Models (LLMs) have shown impressive capabilities, yet they struggle with math reasoning.
Approach:	They propose a coarse-to-fine pruner that prunes unimportant tokens to fit the context window.
Outcome:	The proposed approach outperforms prompting baselines across various LLMs and 5 math datasets and achieves 4.55% absolute improvements without any fine-tuning.

Copied to clipboard

Zhiheng Xi, Senjie Jin, Yuhao Zhou, Rui Zheng, Songyang Gao, Jia Liu, Tao Gui, Qi Zhang, Xuanjing Huang

Let’s Play Across Cultures: A Large Multilingual, Multicultural Benchmark for Assessing Language Models’ Understanding of Sports (2025.emnlp-main)

Challenge:	Existing prompting methods have been used to enhance multistep reasoning capabilities of large language models, but they have overlooked the potential of formulating higher-quality problems.
Approach:	They propose a method that starts from the problem side and refines problems to be more comprehensible and solvable for models.
Outcome:	The proposed method achieves notable and consistent effectiveness on five reasoning benchmarks across different models.

Copied to clipboard

Punit Kumar Singh, Nishant Kumar, Akash Ghosh, Kunal Pasad, Khushi Soni, Manisha Jaishwal, Sriparna Saha, Syukron Abu Ishaq Alfarozi, Asres Temam Abagissa, Kitsuchart Pasupa, Haiqin Yang, Jose G Moreno

Leveraging LLM Reasoning Enhances Personalized Recommender Systems (2024.findings-acl)

Challenge:	Language Models (LMs) are primarily evaluated on globally popular sports, often overlooking regional and indigenous sporting traditions.
Approach:	They propose to use multiple-choice questions (MCQs) to assess LMs' understanding of traditional sports across 60 countries and 6 continents.
Outcome:	The new benchmark will be publicly available, fostering research in culturally aware AI systems.

Copied to clipboard

Alicia Tsai, Adam Kraft, Long Jin, Chenwei Cai, Anahita Hosseini, Taibai Xu, Zemin Zhang, Lichan Hong, Ed H. Chi, Xinyang Yi

Unveiling the Key Factors for Distilling Chain-of-Thought Reasoning (2025.findings-acl)

Challenge:	Recent advances have showcased the potential of Large Language Models (LLMs) in executing reasoning tasks, particularly facilitated by Chain-of-Thought (CoT) prompting.
Approach:	They propose to use Large Language Models to perform tasks with subjectivity and personalized preferences as inputs to RecSys.
Outcome:	The proposed framework aligns with real human judgment on the coherence and faithfulness of LLM reasoning responses.

Copied to clipboard

Xinghao Chen, Zhijing Sun, Guo Wenjin, Miaoran Zhang, Yanjun Chen, Yirong Sun, Hui Su, Yijie Pan, Dietrich Klakow, Wenjie Li, Xiaoyu Shen

The CoT Collection: Improving Zero-shot and Few-shot Learning of Language Models via Chain-of-Thought Fine-Tuning (2023.emnlp-main)

Challenge:	Large Language Models (LLMs) excel in reasoning tasks through Chain-of-Thought prompting.
Approach:	They examine the factors influencing CoT distillation including granularity, format and teacher model.
Outcome:	The proposed model is based on four teacher models and seven student models across seven mathematical and commonsense reasoning datasets.

Copied to clipboard

Seungone Kim, Se Joo, Doyoung Kim, Joel Jang, Seonghyeon Ye, Jamin Shin, Minjoon Seo

Let’s Be Self-generated via Step by Step: A Curriculum Learning Approach to Automated Reasoning with Large Language Models (2025.findings-acl)

Challenge:	Language models with less than 100B parameters perform poorly on chain-of-thought reasoning . we aim to equip smaller LMs with the step-by-step reasoning capability .
Approach:	They propose to equip smaller LMs with the step-by-step reasoning capability by tuning with CoT rationales.
Outcome:	The proposed dataset outperforms large LMs on 4 domain-specific tasks even with demonstrations .

Copied to clipboard

Kangyang Luo, Zichen Ding, Zhenmin Weng, Lingfeng Qiao, Meng Zhao, Xiang Li, Di Yin, Jinlong Shu

ETR: Entropy Trend Reward for Efficient Chain-of-Thought Reasoning (2026.acl-long)

Challenge:	Existing efforts to improve CoT prompting have limitations that require extensive human effort or performance needs to be improved.
Approach:	They propose a prompt approach for automatic reasoning called LBS3 inspired by curriculum learning which better reflects human learning habits.
Outcome:	The proposed approach achieves strongly competitive performance compared to baselines in reasoning-intensive tasks with varying open- and closed-source LLMs.

Copied to clipboard

Xuan Xiong, Huan Liu, Li Gu, Zhixiang Chi, Yue Qiu, Yuanhao YU, Yang Wang

Large Language Models and Multimodal Retrieval for Visual Word Sense Disambiguation (2023.emnlp-main)

Challenge:	Existing methods to shorten CoTs use length penalties or global entropy reduction . Existing approaches to CoT reasoning have significant practical drawbacks .
Approach:	They propose a method that shortens CoTs by length penalties or global entropy reduction . they integrate ETR into Group Relative Policy Optimization and evaluate it .
Outcome:	The proposed objective improves accuracy–efficiency trade-off by +9.9% while reducing CoT length by 67% across four benchmarks.

Copied to clipboard

Anastasia Kritharoula, Maria Lymperaiou, Giorgos Stamou

Automatic Prompt Augmentation and Selection with Chain-of-Thought from Labeled Data (2023.findings-emnlp)

Challenge:	Visual word sense disambiguation (VWSD) is a challenging task involving multiple candidates . context given for an ambiguous word is minimal, most often limited to a single word .
Approach:	They propose to use large language models to enhance given phrases and resolve ambiguity related to the target word.
Outcome:	The proposed frameworks improve the image representation of ambiguous words among candidates and achieve competitive ranking results.

Copied to clipboard

Kashun Shum, Shizhe Diao, Tong Zhang

CR-LLM: A Dataset and Optimization for Concept Reasoning of Large Language Models (2024.findings-acl)

Challenge:	Chain-of-thought (CoT) prompting is a new approach to prompt large language models (LLMs) but most studies rely on human-annotated rational chains to prompt LLMs .
Approach:	They propose a method that augments rational chains from a small labeled dataset and pruning low-quality chains to construct a pool of machine generated rationale chains based on the labels.
Outcome:	The proposed method can bypass human engineering of CoT by automatically augmenting rational chains from a small labeled dataset, and pruning low-quality chains to construct a candidate pool of machine generated rationale chains based on the labels.

Copied to clipboard

Nianqi Li, Jingping Liu, Sihang Jiang, Haiyun Jiang, Yanghua Xiao, Jiaqing Liang, Zujie Liang, Feng Wei, Jinglei Chen, Zhenghong Hao, Bing Han

Towards Better Understanding of Program-of-Thought Reasoning in Cross-Lingual and Multilingual Environments (2025.findings-acl)

Challenge:	Existing concept reasoning related datasets suffer from modeledge leakage and context leakage.
Approach:	They propose a concept reasoning for large language models with modeledge leakage prevention and context leakage preventive methods to improve the models' conceptual reasoning abilities.
Outcome:	The proposed method significantly improves the existing models and reasoning methods, achieving a 7% increase in accuracy compared to CoT and showing better granularity.

Copied to clipboard

Patomporn Payoungkhamdee, Pume Tuchinda, Jinheon Baek, Samuel Cahyawijaya, Can Udomcharoenchaikit, Potsawee Manakul, Peerat Limkonchotiwat, Ekapol Chuangsuwanich, Sarana Nutanong

Enhancing Chain-of-Thought Reasoning via Neuron Activation Differential Analysis (2025.emnlp-main)

Challenge:	Multi-step reasoning is essential for large language models, yet multilingual performance remains challenging.
Approach:	They propose a framework to evaluate Program-of-Thought (PoT) prompting by separating multilingual reasoning from code execution to examine impact of fine-tuning on question-reasoning alignment and reasoning quality.
Outcome:	The proposed framework outperforms CoT fine-tuned models in multilingual settings and shows strong correlation between reasoning quality and answer accuracy.

Copied to clipboard

Yiru Tang, Kun Zhou, Yingqian Min, Xin Zhao, Jing Sha, Zhichao Sheng, Shijin Wang

Agent-GWO: Collaborative Agents for Dynamic Prompt Optimization in Large Language Models (2026.findings-acl)

Challenge:	Existing studies focus on optimizing external components of CoT, but lack internal explanations for the quality of the model's outputs.
Approach:	They propose an efficient method to identify reasoning-critical neurons by analyzing their activation patterns under reasoning chains of varying quality.
Outcome:	The proposed method shows that neurons in the feed-forward layers are critical in the generation of high-quality reasoning chains.

Copied to clipboard

Xudong Wang, Chaoning Zhang, Chenghao Li, Shuxu Chen, Qigan Sun, Jiaquan Zhang, Fachrina Dewi Puspitasari, Tae-Ho Kim, Jiwei Wei, Malu Zhang, Guoqing Wang, Yang Yang, Heng Tao Shen

Large Language Models Are Reasoning Teachers (2023.acl-long)

Challenge:	Existing automatic prompt optimization methods fail to optimize prompts and decoding hyperparameters within a unified framework to achieve stable global improvements.
Approach:	They propose a dynamic prompt optimization framework for complex reasoning that unifies prompt templates and decodes hyperparameters as inheritable agent configurations.
Outcome:	Experiments on multiple mathematical and hybrid reasoning benchmarks show that Agent-GWO improves accuracy and stability over existing prompt optimization methods.

Copied to clipboard

Namgyu Ho, Laura Schmid, Se-Young Yun

CADMate: Generating CAD Assembly Plan with Geometric Chain-of-Thought and Spatial Physical Rewards (2026.acl-long)

Challenge:	Recent studies have shown that chain-of-thought (CoT) prompting can elicit language models to solve complex reasoning tasks step-by-step.
Approach:	They propose a method that uses large model samples as reasoning teachers to fine-tune smaller models.
Outcome:	The proposed method outperforms prompt-based methods and the teacher model in many tasks and extends it by leveraging the teacher's ability to generate multiple rationales for each original sample.

Copied to clipboard

Jiali Chen, DingBa Fu, Xusen Hei, Yuhang Liu, Yiyang Chen, Jiayuan Xie, Wenqi Fan, Yi Cai

Probabilistic Tree-of-thought Reasoning for Answering Knowledge-intensive Complex Questions (2023.findings-emnlp)

Challenge:	Computer-aided design (CAD) is crucial in prototyping complex 3D objects . designers manually define assembly sequences for individual CAD parts .
Approach:	They propose a framework for computer-aided design that predicts actions for CAD parts . they use a reference design image and disassembled parts to generate 6-DoF transformations .
Outcome:	The proposed framework outperforms existing MLLMs in the design of CAD assemblies.

Copied to clipboard

Shulin Cao, Jiajie Zhang, Jiaxin Shi, Xin Lv, Zijun Yao, Qi Tian, Lei Hou, Juanzi Li

HalluMeasure: Fine-grained Hallucination Measurement Using Chain-of-Thought Reasoning (2024.emnlp-main)

Challenge:	Large language models (LLMs) are capable of answering knowledge-intensive complex questions with chain-of-thought reasoning.
Approach:	They propose a method to solve complex questions with a tree-of-thought approach using parametric knowledge and retrieved external knowledge to augment CoT reasoning.
Outcome:	The proposed approach outperforms SOTA methods on three Complex QA datasets under the open-domain setting.

Copied to clipboard

Shayan Ali Akbar, Md Mosharaf Hossain, Tess Wood, Si-Chi Chin, Erica Salinas, Victor Alvarez, Erwin Cornejo

Style-Compress: An LLM-Based Prompt Compression Framework Considering Task-Specific Styles (2024.findings-emnlp)

Challenge:	HalluMeasure is a new LLM-based hallucination detection mechanism that decomposes an LLM response into atomic claims and evaluates each claim against the provided reference context.
Approach:	They propose a new LLM-based hallucination detection mechanism that decomposes an LLM response into atomic claims and evaluates each atomic claim against the provided reference context.
Outcome:	The proposed model can detect 3 major categories of hallucinations and 10 more specific subtypes which help to identify reasons behind the hallucinian errors.

Copied to clipboard

Xiao Pu, Tianxing He, Xiaojun Wan

Logical Phase Transitions: Understanding Collapse in LLM Logical Reasoning (2026.acl-long)

Challenge:	Prompt compression reduces inference time and costs while maintaining informativeness for different usage scenarios.
Approach:	They propose a framework that adapts a smaller language model to compress prompts for a larger model on a new task without additional training.
Outcome:	The proposed framework outperforms two baseline models in four tasks . iteratively generates and selects effective compressed prompts as task-specific demonstrations .

Copied to clipboard

Xinglang Zhang, Yunyao Zhang, ZeLiang Chen, Junqing Yu, Wei Yang, Zikai Song

ThinkEdit: Interpretable Weight Editing to Mitigate Overly Short Thinking in Reasoning Models (2025.emnlp-main)

Challenge:	Symbolic logical reasoning is a critical yet underexplored capability of large language models (LLMs).
Approach:	They propose a framework that aligns natural language with logical symbols to establish a shared representation and reshapes training dynamics around phase-transition boundaries to progressively strengthen reasoning at increasing logical depths.
Outcome:	The proposed framework mitigates logical reasoning collapse at high complexity while improving generalization to unseen logical compositions.

Copied to clipboard

Chung-En Sun, Ge Yan, Tsui-Wei Weng

Stepwise Reasoning Checkpoint Analysis: A Test Time Scaling Method to Enhance LLMs’ Reasoning (2025.emnlp-main)

Challenge:	Recent studies have shown that Large Language Models (LLMs) augmented with chain-of-thought (CoT) reasoning demonstrate impressive problem-solving abilities.
Approach:	They propose a weight-editing approach to reduce overly short reasoning by steering the model along a linear direction in the representation space.
Outcome:	The proposed model reduces overly short reasoning and yields significant accuracy gains on multiple math benchmarks.

Copied to clipboard

Zezhong Wang, Xingshan Zeng, Weiwen Liu, Yufei Wang, Liangyou Li, Yasheng Wang, Lifeng Shang, Xin Jiang, Qun Liu, Kam-Fai Wong

Language Models Can Easily Learn to Reason from Demonstrations (2025.findings-emnlp)

Challenge:	Existing methods that use Chain-of-Thought suffer from path homogenization and inefficient use of intermediate results.
Approach:	They propose a framework that introduces checkpoints between reasoning steps to reduce path homogenization and create fault-tolerant mechanisms.
Outcome:	The proposed framework reduces path homogenization and creates fault-tolerant mechanism by utilizing high-quality intermediate results.

Copied to clipboard

Dacheng Li, Shiyi Cao, Tyler Griggs, Shu Liu, Xiangxi Mo, Eric Tang, Sumanth Hegde, Kourosh Hakhamaneshi, Shishir G Patil, Matei Zaharia, Joseph E. Gonzalez, Ion Stoica

Revisiting Relation Extraction in the era of Large Language Models (2023.acl-long)

Challenge:	Large reasoning models (LRMs) tackle complex problems by following long chain-of-thoughts (Long CoT) however, the training techniques and data requirements to elicit Long CoT remain poorly understood.
Approach:	They propose to use data-efficient supervised fine-tuning and parameter-efficient low-rank adaptation to elicit Long CoT reasoning.
Outcome:	The proposed model can learn Long CoT reasoning through data-efficient supervised fine-tuning and parameter-efficient low-rank adaptation.

Copied to clipboard

Somin Wadhwa, Silvio Amir, Byron Wallace

Re-Reading Improves Reasoning in Large Language Models (2024.emnlp-main)

Challenge:	Standard supervised approaches to RE learn to tag tokens comprising entity spans and then predict the relationship between them.
Approach:	They propose to use large language models for RE to evaluate their performance . they use GPT-3 and Flan-T5 large to train RE .
Outcome:	The proposed model outperforms existing models on a sequence-to-sequence task under varying levels of supervision.

Copied to clipboard

Xiaohan Xu, Chongyang Tao, Tao Shen, Can Xu, Hongbo Xu, Guodong Long, Jian-Guang Lou, Shuai Ma

Hallucination Detection in Structured Query Generation via LLM Self-Debating (2025.findings-emnlp)

Challenge:	Unlike thought-eliciting prompting methods, RE2 shifts the focus to the input by processing questions twice, thereby enhancing the understanding process.
Approach:	They introduce a simple, yet general and effective prompting method, RE2, which rereads the question as input.
Outcome:	The proposed method demonstrates strong generality and compatibility with most thought-eliciting prompting methods, including CoT.

Copied to clipboard

Miaoran Li, Jiangning Chen, Minghua Xu, Xiaolong Wang

Preemptive Answer “Attacks” on Chain-of-Thought Reasoning (2024.findings-acl)

Challenge:	Hallucination remains a key challenge in applying large language models to structured query generation . we propose the Self-Debating framework to enhance detection performance .
Approach:	They propose a framework that prompts an LLM to generate contrastive explanations from opposing perspectives . they also propose 'self-debating' framework to enhance detection performance .
Outcome:	The proposed framework outperforms LLM-as-a-Judge baselines in hallucination detection . the framework generates contrastive explanations from opposing perspectives .

Copied to clipboard

Rongwu Xu, Zehan Qi, Wei Xu

EvolvR: Self-Evolving Pairwise Reasoning for Story Evaluation to Enhance Generation (2026.acl-long)

Challenge:	Large language models (LLMs) have impressive reasoning capabilities when coupled with Chain-of-Thought (CoT) prompting.
Approach:	They propose a scenario where the LLM obtains an answer before engaging in reasoning, and propose two measures to bolster the robustness of this approach.
Outcome:	The proposed model significantly impairs its reasoning capability across various CoT methods and a broad spectrum of datasets.

Copied to clipboard

Xinda Wang, Zhengxu Hou, Yangshijie Zhang, null Yanbingren, Jialin Liu, ChenZhuo Zhao, Zhibo Yang, Bin-Bin Yang, Feng Xiao

LangSuit·E: Planning, Controlling and Interacting with Large Language Models in Embodied Text Environments (2024.findings-acl)

Challenge:	Existing methods for story evaluation lack reasoning capabilities for open-source models . evolvR framework provides high-fidelity evaluators for story generation tasks .
Approach:	They propose a framework that self-synthesizes chain-of-thought data via a multi-persona strategy . they propose evolvR to provide a reward model for story generation .
Outcome:	The proposed framework achieves state-of-the-art performance on three evaluation benchmarks . it also enhances the quality of generated stories, validating the superiority of the framework .

Copied to clipboard

Zixia Jia, Mengmeng Wang, Baichen Tong, Song-Chun Zhu, Zilong Zheng

Causal-ESC: Reliable Policy Learning for Emotional Support Conversation via Causal Inference (2026.acl-long)

Challenge:	Recent advances in Large Language Models (LLMs) have shown inspiring achievements in constructing autonomous agents that rely on language descriptions as inputs.
Approach:	They propose a flexible and simulation-free testbed that simulates 6 representative embodied tasks in textual embodies.
Outcome:	The proposed testbed offers adaptability to diverse environments without multiple simulation engines and allows easy customization of communication and action strategies.

Copied to clipboard

Xv Wang, Wang Zhenyu, Guanyu Zheng, Rui Zhang

Dialogue Systems & Conversational Agents

Challenge:	Existing approaches to Emotional Support Conversation (ESC) are mechanistically opaque and lacks a causal mechanism between dialogue features and effective empathic strategies.
Approach:	They propose a framework that uses Doubly Robust learning to model causal effects of utterance features on strategy selection.
Outcome:	The proposed framework outperforms state-of-the-art baselines in empathy and helpfulness and provides a theoretically grounded, interpretable solution to the mechanistic interpretability dilemma in affective computing.

Understanding the Language Model to Solve the Symbolic Multi-Step Reasoning Problem from the Perspective of Buffer Mechanism (2025.findings-emnlp)

Copied to clipboard

Zhiwei Wang, Yunji Wang, Zhongwang Zhang, Zhangchen Zhou, Hui Jin, Tianyang Hu, Jiacheng Sun, Zhenguo Li, Yaoyu Zhang, Zhi-Qin John Xu

Mapping the Minds of LLMs: A Graph-Based Analysis of Reasoning LLMs (2025.emnlp-main)

Challenge:	Large language models struggle with complex reasoning tasks, such as mathematical problem-solving.
Approach:	They constructed a symbolic multi-step reasoning task to investigate the information propagation mechanisms in Transformer models when solving the task through direct answering and Chain-of-Thought (CoT) reasoning.
Outcome:	The proposed algorithm improves on 7 multi-step reasoning datasets, while introducing only 132 trainable parameters.

Copied to clipboard

Zhen Xiong, Yujun Cai, Zhecheng Li, Yiwei Wang

AdaTooler-V: Adaptive Tool-Use for Images and Videos (2026.findings-acl)

Challenge:	Large Reasoning Models (LRMs) often display unstable behaviors, e.g., hallucinating unsupported premises, overthinking simple tasks, and displaying higher sensitivity to prompt variations.
Approach:	They propose a graph-based analytical framework that clusters long, verbose CoT outputs into semantically coherent reasoning steps, then constructs directed reasoning graphs to capture contextual and logical dependencies among these steps.
Outcome:	The proposed framework enables quantitative evaluation of internal reasoning structure and quality beyond conventional metrics and provides practical insights for prompt engineering and cognitive analysis of LLMs.

Copied to clipboard

Chaoyang Wang, Kaituo Feng, Dongyang Chen, Zhongyu Wang, Zhixun Li, Sicheng Gao, Meng Meng, Xu Zhou, Manyuan Zhang, Yuzhang Shang, Xiangyu Yue

Explainable Chain-of-Thought Reasoning: An Empirical Analysis on State-Aware Reasoning Dynamics (2025.findings-emnlp)

Challenge:	Existing models exhibit blind tool-use reasoning patterns, which significantly increases inference overhead and degrades model performance.
Approach:	They propose an MLLM that performs adaptive tool-use by determining whether a visual problem truly requires tools.
Outcome:	The proposed model outperforms existing methods in visual reasoning tasks.

Copied to clipboard

Sheldon Yu, Yuxin Xiong, Junda Wu, Xintong Li, Tong Yu, Xiang Chen, Ritwik Sinha, Jingbo Shang, Julian McAuley

Answer Convergence as a Signal for Early Stopping in Reasoning (2025.emnlp-main)

Challenge:	Recent advances in chain-of-thought prompting have demonstrated the ability of large language models to perform multi-step reasoning.
Approach:	They propose a framework to analyze latent dynamics of CoT trajectories for interpretability . they segment generated CoT into discrete reasoning steps and abstract each step into a spectral embedding based on token-level Gram matrices .
Outcome:	The proposed framework segments generated CoT steps into discrete reasoning steps, abstracts each step into a spectral embedding based on token-level Gram matrices, and clusters these embeddements into semantically meaningful latent states.

Copied to clipboard

Xin Liu, Lu Wang

DVMap: Fine-Grained Pluralistic Value Alignment via High-Consensus Demographic-Value Mapping (2026.acl-long)

Challenge:	a systematic study suggests that chain-of-thought prompting is unnecessary for producing correct answers.
Approach:	They propose three inference-time strategies to improve model efficiency by boosting end-of-reasoning signals and early stopping . they propose a method that learns when to stop based on internal activations .
Outcome:	The proposed methods reduce token usage with little or no accuracy drop on natural questions . the proposed methods also reduce tokens by over 40% on naturalquestions .

Copied to clipboard

Pengyun Zhu, Yuqi Ren, Zhen Wang, Lei Yang, Deyi Xiong

Challenge:	Current Large Language Models (LLMs) rely on coarse-grained national labels for pluralistic value alignment.
Approach:	They propose a framework for fine-grained pluralistic value alignment using demographic constraints.
Outcome:	The proposed framework can identify groups with predictable, high-consensus value preference . it achieves 48.6% accuracy, surpassing open-source LLM DeepSeek-v3.2 .

Chain-of-Thought Tuning: Masked Language Models can also Think Step By Step in Natural Language Understanding (2023.emnlp-main)

Copied to clipboard

Caoyun Fan, Jidong Tian, Yitian Li, Wenqing Chen, Hao He, Yaohui Jin

Unilaw-R1: A Large Language Model for Legal Reasoning with Reinforcement Learning and Iterative Inference (2025.emnlp-main)

Challenge:	Chain-of-Thought (CoT) is a technique that guides large language models to decompose complex tasks into multi-step reasoning processes.
Approach:	They propose a two-step reasoning framework based on prompt tuning to implement step-by-step thinking for MLMs on NLU tasks.
Outcome:	The proposed framework outperforms baselines and achieves state-of-the-art performance on two NLU tasks.

Copied to clipboard

Hua Cai, Shuang Zhao, Liang Zhang, Xuli Shen, Qing Xu, Weilin Shen, Zihao Wen, Tianke Ban

Beyond Static Testbeds: An Interaction-Centric Agent Simulation Platform for Dynamic Recommender Systems (2025.emnlp-main)

Challenge:	Reasoning-focused large language models (LLMs) are rapidly evolving across various domains, yet their capabilities in handling complex legal problems remain underexplored.
Approach:	They propose a large language model tailored for legal reasoning with a 7-billion parameter scale and a two-stage training strategy combining Supervised Fine-Tuning and Reinforcement Learning.
Outcome:	The proposed model outperforms all models of similar scale on authoritative benchmarks and outperformed Qwen-2.5-7B-Instruct (46.6%) by an average margin of 6.6%.

Copied to clipboard

Song Jin, Juntian Zhang, Yuhan Liu, Xun Zhang, Yufei Zhang, Guojun Yin, Fei Jiang, Wei Lin, Rui Yan

Challenge:	Existing platforms lack a mechanism for user actions to dynamically reshape the environment.
Approach:	They propose a novel agent-based simulation platform for recommender systems with a robust interaction mechanism.
Outcome:	The proposed platform improves the credibility of the simulation and replicates the Matthew Effect and Brand Loyalty.

Stepwise Perplexity-Guided Refinement for Efficient Chain-of-Thought Reasoning in Large Language Models (2025.findings-acl)

Copied to clipboard

Yingqian Cui, Pengfei He, Jingying Zeng, Hui Liu, Xianfeng Tang, Zhenwei Dai, Yan Han, Chen Luo, Jing Huang, Zhen Li, Suhang Wang, Yue Xing, Jiliang Tang, Qi He

Mentor-KD: Making Small Language Models Better Multi-step Reasoners (2024.emnlp-main)

Challenge:	Chain-of-Thought (CoT) reasoning has improved the performance of large language models (LLMs) however, the detailed reasoning process in CoT often incurs long generation times and high computational costs due to the inclusion of unnecessary steps.
Approach:	They propose a method to identify critical reasoning steps using perplexity as a measure of their importance.
Outcome:	The proposed method achieves a better balance between reasoning accuracy and efficiency of CoT.

Copied to clipboard

Hojae Lee, Junho Kim, SangKeun Lee

Vulnerability of LLMs to Vertically Aligned Text Manipulations (2025.acl-long)

Challenge:	Large Language Models (LLMs) have shown impressive emergent capabilities by leveraging Chain-of-Thought (CoT) prompting.
Approach:	They propose a Knowledge Distillation approach which transfers multi-step reasoning ability of Large Language Models (LLMs) to smaller LMs by fine-tuning language models of multi- step rationales generated by LLM teachers.
Outcome:	The proposed method is able to transfer multi-step reasoning ability of LLMs to smaller LMs while addressing data quality and soft label provision.

Copied to clipboard

Zhecheng Li, Yiwei Wang, Bryan Hooi, Yujun Cai, Zhen Xiong, Nanyun Peng, Kai-Wei Chang

Text Classification

Challenge:	Recent research shows that vertical text input significantly degrades the accuracy of large language models (LLMs) in text classification tasks.
Approach:	They investigate the impact of vertical text input on the performance of LLMs . they find that chain of thought reasoning does not help LLM recognize vertical input .
Outcome:	The proposed model can significantly mislead models, posing a risk of bypassing detection in real-world scenarios involving harmful or sensitive information.

Optimizing Reasoning for Text-to-SQL with Execution Feedback (2025.findings-acl)

Copied to clipboard

Bohan Zhai, Canwen Xu, Yuxiong He, Zhewei Yao

X-Router: Decoupling Knowledge and Reasoning for Cost-Effective LLM Inference (2026.findings-acl)

Challenge:	Large language models excel in many reasoning tasks, but their ability to leverage Chain-of-Thought (CoT) reasoning remains underexplored.
Approach:	They propose a framework that iteratively optimizes open-source LLMs by combining CoT reasoning with off-policy and on-poly DPO, relying solely on execution accuracy as feedback.
Outcome:	The proposed framework improves execution accuracy on BIRD and Spider datasets.

Copied to clipboard

Zixuan Wang, Yinze Ding, Zihan Wang, Jinyu Guo, Zhenhong Zhou, Junhao Dong, Chaomeng Chen

Direct Behavior Optimization: Unlocking the Potential of Lightweight LLMs (2025.findings-acl)

Challenge:	Existing adaptive methods focus on a single axis, overlooking evidence need and reasoning depth are only partially correlated.
Approach:	They propose a dual-axis routing framework that separates retrieval necessity from reasoning necessity under a user-defined cost–quality trade-off.
Outcome:	The proposed framework reduces token usage and latency while improving answer quality over strong baselines.

Copied to clipboard

Hongming Yang, Shi Lin, Jun Shao, Changting Lin, Donghai Zhu, Meng Han, Qinglei Kong

Can LLMs Help Uncover Insights about LLMs? A Large-Scale, Evolving Literature Analysis of Frontier LLMs (2025.acl-long)

Challenge:	Existing prompt optimization methods rely on extensive manual effort or meta-cognitive abilities, making them less effective for LwLLMs.
Approach:	They propose a direct behavior optimization parameter that transforms the optimization of complex prompts into discrete, quantifiable execution sequences using a gradient-free Monte Carlo Tree Search.
Outcome:	The proposed method outperforms current prompt optimization methods on seven challenging tasks where state-of-the-art LLMs excel but LwLLMs generally underperform.

Copied to clipboard

Jungsoo Park, Junmo Kang, Gabriel Stanovsky, Alan Ritter

Tree of Problems: Improving structured problem solving with compositionality (2024.emnlp-main)

Challenge:	Recent surveys of literature highlight the overwhelming growth of Large Language Models (LLMs).
Approach:	They propose a semi-automated literature analysis approach that automates literature analysis using LLMs.
Outcome:	The proposed approach reduces paper surveying and data extraction by 93% compared to manual methods.

Copied to clipboard

Armel Zebaze, Benoît Sagot, Rachel Bawden

Does Reasoning Introduce Bias? A Study of Social Bias Evaluation and Mitigation in LLM Reasoning (2025.findings-emnlp)

Challenge:	Large Language Models (LLMs) have demonstrated remarkable performance across multipletasks through in-context learning.
Approach:	They propose a Tree of Problems (ToP) that is a simpler version of Tree of Thoughts (toT) they propose 'in-context learning' is the ability of Large Language Models (LLMs) to perform a task with the help of a few demonstrations within their context.
Outcome:	The proposed approach outperforms ToT and GoT and performs better on complex reasoning tasks.

Copied to clipboard

Xuyang Wu, Jinming Nian, Ting-Ruen Wei, Zhiqiang Tao, Hsin-Tai Wu, Yi Fang

On the Same Wavelength? Evaluating Pragmatic Reasoning in Language Models across Broad Concepts (2025.emnlp-main)

Challenge:	Recent advances in large language models have enabled automatic generation of chain-of-thought reasoning . however, when reasoning steps reflect social stereotypes, they can reinforce harmful associations and lead to misleading conclusions.
Approach:	They propose a method that detects how model predictions change across incremental reasoning steps.
Outcome:	The proposed method outperforms a stereotype-free baseline and improves accuracy.

Copied to clipboard

Linlu Qiu, Cedegao E. Zhang, Joshua B. Tenenbaum, Yoon Kim, Roger P. Levy

SciVerse: Unveiling the Knowledge Comprehension and Visual Reasoning of LMMs on Multi-modal Scientific Problems (2025.findings-acl)

Challenge:	Language models (LMs) are increasingly used as conversational agents because of their pragmatic reasoning abilities.
Approach:	They propose an evaluation framework derived from Wavelength, a popular communication game where a speaker and a listener communicate about a broad range of concepts in a granular manner.
Outcome:	The proposed evaluation framework outperforms direct and Chain-of-Thought (CoT) prompting on language comprehension and language production tasks.

Copied to clipboard

Ziyu Guo, Renrui Zhang, Hao Chen, Jialin Gao, Dongzhi Jiang, Jiaze Wang, Pheng-Ann Heng

FACT-E: Causality-Inspired Evaluation for Trustworthy Chain-of-Thought Reasoning (2026.findings-acl)

Challenge:	SciVerse is a multi-modal scientific evaluation benchmark to assess large multi-models . it examines the scientific knowledge comprehension, multi-mod content interpretation and Chain-of-Thought reasoning . authors examine the scientific proficiency of LMMs in scientific domains based on their work .
Approach:	They propose a multi-modal scientific evaluation benchmark to thoroughly assess Large Multi-modal Models across 5,735 test instances in five different versions.
Outcome:	The proposed evaluation reveals critical limitations in LMMs' scientific proficiency and provides new insights into future developments.

Copied to clipboard

Yuxi Sun, Aoqi Zuo, Haotian Xie, Wei Gao, Mingming Gong, Jing Ma

CoMAT: Chain of Mathematically Annotated Thought Improves Mathematical Reasoning (2025.emnlp-main)

Challenge:	Existing models generate explanations that appear coherent while containing unfaithful intermediate steps.
Approach:	They propose a causality-inspired framework for evaluating CoT quality using controlled perturbations as an instrumental signal to separate genuine step-to-step dependence from bias-driven artifacts.
Outcome:	Experiments on GSM8K, MATH, and CommonsenseQA show that FACT-E improves reasoning-trajectory selection and yields stronger in-context learning exemplars.

Copied to clipboard

Joshua Ong Jun Leang, Aryo Pradipta Gema, Shay B Cohen

Uncovering the Impact of Chain-of-Thought Reasoning for Direct Preference Optimization: Lessons from Text-to-SQL (2025.acl-long)

Challenge:	Mathematical reasoning remains a significant challenge for large language models (LLMs), despite advances in prompting techniques such as Chain-of-Thought (CoT).
Approach:	They propose a framework that enhances reasoning through two stages: Symbolic Conversion and Reasoning Execution.
Outcome:	The proposed framework outperforms traditional CoT on six out of seven benchmarks across four LLMs.

Copied to clipboard

Hanbing Liu, Haoyang Li, Xiaokang Zhang, Ruotong Chen, Haiyong Xu, Tian Tian, Qi Qi, Jing Zhang

How Do Large Vision-Language Models See Text in Image? Unveiling the Distinctive Role of OCR Heads (2025.emnlp-main)

Challenge:	Direct Preference Optimization (DPO) is effective in complex reasoning tasks like math word problems and code generation, but Text-to-SQL datasets often include only final answers (gold SQL queries) without detailed CoT solutions.
Approach:	They found that Direct Preference Optimization (DPO) is crucial for unlocking DPO's potential by augmenting Text-to-SQL datasets with synthetic CoT solutions.
Outcome:	The proposed method achieves consistent and significant performance improvements on Text-to-SQL datasets.

Copied to clipboard

Ingeol Baek, Hwan Chang, Sunghyun Ryu, Hwanhee Lee

LogicTree: Structured Proof Exploration for Coherent and Rigorous Logical Reasoning with Large Language Models (2025.emnlp-main)

Challenge:	Despite advances in Large Vision Language Models, a gap remains in their interpretability and performance.
Approach:	They identify the Optical Character Recognition Head (OCR Head) heads that are more efficient at recognizing text from images.
Outcome:	The Optical Character Recognition Head (OCR Head) is identified as the most efficient head for recognizing text from images.

Copied to clipboard

Kang He, Kaushik Roy

Step Guided Reasoning: Improving Mathematical Reasoning using Guidance Generation and Step Reasoning (2025.emnlp-main)

Challenge:	Large language models (LLMs) have remarkable multi-step reasoning capabilities, but they still face challenges in complex logical reasoning.
Approach:	They propose an algorithm-guided search framework that automates structured proof exploration and ensures logical coherence.
Outcome:	The proposed framework outperforms o3-mini and chain-of-thought with average gains of 23.6% and 12.5% on five datasets.

Copied to clipboard

Lang Cao, Yingtian Zou, Chao Peng, Renhong Chen, Wu Ning, Yitong Li

TOXIFRENCH: Benchmarking and Enhancing Language Models via CoT Fine-Tuning for French Toxicity Detection (2026.findings-acl)

Challenge:	Existing approaches to improve mathematical reasoning require extensive datasets for training or depend on few-shot methods that compromise computational accuracy.
Approach:	They propose a training-free adaptation framework that efficiently equips general-purpose pre-trained language models with enhanced mathematical reasoning capabilities.
Outcome:	The proposed framework outperforms Qwen2.5-72B-Math-Instruct on MMLU-STEM with a score of 90.9%, compared to 87.3%.

Copied to clipboard

Axel Delaval, Shujian Yang, Haicheng Wang, Han Qiu, Jialiang LU

When Is Thinking Enough? Early Exit via Sufficiency Assessment for Efficient Reasoning (2026.acl-long)

Challenge:	toxicity detection in French remains underdeveloped due to the lack of culturally relevant, human-annotated, large-scale datasets.
Approach:	They propose a method that generalizes French online comments using a semi-automated annotation pipeline that reduces manual labeling to only 10% through high-confidence LLM-based pre-annotation and human verification.
Outcome:	The proposed model outperforms GPT-4o and DeepSeek-R1 on the benchmark while maintaining cross-lingual capabilities.

Copied to clipboard

Yang Xiang, Yixin Ji, Ruotao Xu, Dan Qiao, Zheming Yang, Juntao Li, Min Zhang

ThinkBrake: Efficient Reasoning via Log-Probability Margin Guided Decoding (2026.findings-acl)

Challenge:	Existing approaches to early exit reasoning often rely on handcrafted or empirical indicators that are unreliable and impractical.
Approach:	They propose a framework that allows LRMs to assess the sufficiency of its chain-of-thought and determine the optimal point for early exit.
Outcome:	The proposed framework reduces reasoning length by 28.9%–34.9% with minimal performance loss, effectively mitigating overthinking.

Copied to clipboard

Sangjun Song, Minjae Oh, Seungkyu Lee, Sungmin Jo, Yohan Jo

Decoupling the Effect of Chain-of-Thought Reasoning: A Human Label Variation Perspective (2026.findings-acl)

Challenge:	Recent advances in Large Reasoning Models (LRMs) have demonstrated remarkable capabilities across various tasks.
Approach:	They propose a system that stops reasoning when the margin between continuation token and lt;/think gt; narrows.
Outcome:	The proposed model reduces thinking token usage by 30% and improves accuracy by 8% while reducing thinking tokens by 72%.

Copied to clipboard

Beiduo Chen, Tiancheng Hu, Caiqi Zhang, Robert Litschko, Anna Korhonen, Barbara Plank

A Reinforcement Learning Framework for Cross-Lingual Stance Detection Using Chain-of-Thought Alignment (2025.findings-acl)

Challenge:	Reasoning-tuned large language models (LLMs) with long Chain-of-Thought excel at single-answer tasks, yet their ability to model Human Label Variation remains underexplored.
Approach:	They conduct systematic disentanglement experiments to isolate the effect of reasoning text from intrinsic model priors on distribution-based tasks.
Outcome:	The proposed model improves distributional alignment, but distributional ranking is governed by model priors.

Copied to clipboard

Binghui Li, Minghui Zou, Xiaowang Zhang, Shizhan Chen, Zhiyong Feng

Cross-Lingual Transfer

Challenge:	Existing approaches to cross-lingual stance detection can't effectively perform cross-linguistic transfer of complex reasoning processes.
Approach:	They propose a framework to facilitate cross-lingual transfer of complex reasoning processes in stance detection by using cross-linguistic Chain-of-Thought alignment to obtain high-quality CoTs generated from target language inputs.
Outcome:	The proposed framework outperforms competing models on four multilingual datasets.

Large Reasoning Models Are (Not Yet) Multilingual Latent Reasoners (2026.findings-acl)

Copied to clipboard

Yihong Liu, Raoyuan Zhao, Hinrich Schuetze, Michael A. Hedderich

Think Less, Know More: State-Aware Reasoning Compression with Knowledge Guidance for Efficient Reasoning (2026.findings-acl)

Challenge:	Recent work shows that large reasoning models arrive at the correct answer before completing textual reasoning steps, indicating the presence of latent reasoning.
Approach:	They conduct a systematic investigation of multilingual latent reasoning in large reasoning models across 11 languages.
Outcome:	The proposed model arrive at the correct answer before completing the reasoning steps, indicating the presence of latent reasoning.

Copied to clipboard

Yi Sui, Chaozhuo Li, Dawei Song

Large Language Models with Temporal Reasoning for Longitudinal Clinical Summarization and Prediction (2025.findings-emnlp)

Challenge:	Existing CoT compression methods struggle to balance accuracy and efficiency . long CoT reasoning also introduces an overthinking phenomenon, authors say .
Approach:	They propose a framework that performs step-wise CoT compression by modeling stage-specific redundancy sources and integrating with a retrieval-augmented guidance.
Outcome:	The proposed framework reduces average response length by 59.9% while improving accuracy by 4.8 points over existing methods.

Copied to clipboard

Maya Kruse, Shiyue Hu, Nicholas Derby, Yifu Wu, Samantha Stonbraker, Bingsheng Yao, Dakuo Wang, Elizabeth M. Goldberg, Yanjun Gao

Adaptive Spatial and Temporal Redundancy Optimization for Efficient Reasoning in Large Language Models (2026.acl-long)

Challenge:	Recent advances in large language models have shown potential in clinical text summarization, but their ability to handle long patient trajectories with multi-modal data spread across time remains underexplored.
Approach:	They evaluate open-source large language models, their Retrieval Augmented Generation variants and chain-of-thought prompting on long-context clinical summarization and prediction.
Outcome:	The proposed models can synthesize structured and unstructured EHR data while reasoning over temporal coherence.

Copied to clipboard

Tianle Chen, Pengyu Cheng, Qiyuan Zhu, Jiacheng Wang, Bei Liu, Hao Gu, Ruijie Shen, Xiaofeng Hou, Sirui Han, Jiacheng Liu

LLM-Guided Semantic Relational Reasoning for Multimodal Intent Recognition (2025.emnlp-main)

Challenge:	Existing research to improve CoT efficiency falls into three categories, each with distinct limitations.
Approach:	They propose a training-free framework that addresses both dimensions of CoT reasoning by applying a progressive precision reduction strategy coupled with an entropy-based confidence mechanism for adaptive termination.
Outcome:	Empirical results show that the proposed framework achieves 11.3 efficiency gain without compromising accuracy.

Copied to clipboard

Qianrui Zhou, Hua Xu, Yifan Wang, Xinzhi Dong, Hanlei Zhang

AAPO: Enhancing the Reasoning Capabilities of LLMs with Advantage Margin (2026.acl-long)

Challenge:	Existing methods for understanding intents from multimodal signals exhibit limitations in their modality-level reliance, constraining relational reasoning over fine-grained semantics for complex intent understanding.
Approach:	They propose a method that harnesses the expansive knowledge of large language models to establish semantic foundations that boost smaller models’ relational reasoning performance.
Outcome:	The proposed method outperforms state-of-the-art methods on multimodal intent and dialogue act recognition tasks and shows consistent performance gains across diverse semantic understanding scenarios.

Copied to clipboard

Jian Xiong, Jingbo Zhou, Jingyong Ye, Qiang Huang, Dejing Dou

Probe Then Retrieve and Reason: Distilling Probing and Reasoning Capabilities into Smaller Language Models (2024.lrec-main)

Challenge:	Reinforcement learning (RL) has emerged as an effective approach for enhancing the reasoning capabilities of large language models.
Approach:	They propose an algorithm that optimizes cross-entropy loss using advantages enhanced through a margin-based estimation scheme.
Outcome:	Experimental results show that AAPO improves group relative advantage estimation compared to other methods.

Copied to clipboard

Yichun Zhao, Shuheng Zhou, Huijia Zhu

TOOLCAD: Exploring Tool-Using Large Language Models in Text-to-CAD Generation with Reinforcement Learning (2026.findings-acl)

Challenge:	Recent research efforts have focused on distilling Large Language Models into Small Language Model (SLMs) however, the results of CoT distillation are inadequate for knowledge-intensive reasoning tasks.
Approach:	They propose a retrieval-based framework which distills question probing and reasoning capabilities from Large Language Models into SLMs.
Outcome:	The proposed framework improves probing and reasoning capabilities of large language models in knowledge-intensive reasoning tasks.

Copied to clipboard

Yifei Gong, Xing Wu, Wenda Liu, null Tukang

Simple Role Assignment is Extraordinarily Effective for Safety Alignment (2026.findings-acl)

Challenge:	Large Language Models (LLMs) have shown remarkable advances in enabling language agents to tackle real-world tasks.
Approach:	They propose a tool-using agent-based CAD framework that automates text-to-CAD modeling . they propose an interactive CAD gym to roll out reasoning and tool-augmented interaction trajectories with the CAD engine .
Outcome:	The proposed framework can generalize across complex modeling tasks, supporting their open-source counterparts.

Copied to clipboard

Zhou Ziheng, Jiakun Ding, Zhaowei Zhang, Ruosen Gao, Ying Nian Wu, Demetri Terzopoulos, Yipeng Kang, Fangwei Zhong, Junqi Wang

Challenge:	a new study proposes a role-conditioned pipeline for value alignment . principles alone are incomplete, and they provide little guidance on when and how a value applies in context.
Approach:	They propose a role-conditioned pipeline with role-based critics and a model-free approach that is based on role conditioning.
Outcome:	The proposed approach outperforms principle-based, Chain-of-Thought and other benchmarks.

Reason Only When Needed: Efficient Generative Reward Modeling via Model-Internal Uncertainty (2026.findings-acl)

Copied to clipboard

Chao Xue, Yao Wang, Mengqiao Liu, Di Liang, Xingsheng Han, Peiyang Liu, Xianjie Wu, Chenyao Lu, Lei Jiang, Yu Lu, Haibo Shi, Shuang Liang, Minlong Peng, Flora D. Salim

Learning to Think on Hypergraph: HyperCoT for Structure-Guided N-ary Knowledge Graph Completion (2026.acl-long)

Challenge:	Existing approaches to generating reward models rely on voting-based mechanisms to evaluate CoT outputs.
Approach:	They propose an efficient generative reward modeling framework grounded in model-internal uncertainty.
Outcome:	The proposed framework reduces inference cost while improving answer accuracy.

Copied to clipboard

Mengxue Yang, Jinming Li, Chun Yang, Jiaqi Zhu, Jiafan Li, Guanhua Zhang, Ying Li

Structured Data in NLP

Challenge:	Existing methods to solve knowledge hypergraph link prediction problem are limited by their ability to generate chain-of-thought (CoT) representations.
Approach:	They propose a structure-aware approach that models multi-hop structural reasoning as a depth-sensitive progressive evidence accumulation process.
Outcome:	Experiments on three real-world datasets show that HyperCoT outperforms strong n-ary KGC baselines while yielding interpretable multi-hop reasoning traces.

TABARD: A Novel Benchmark for Tabular Anomaly Analysis, Reasoning and Detection (2025.findings-emnlp)

Copied to clipboard

Manan Roy Choudhury, Anirudh Iyengar Kaniyar Narayana Iyengar, Shikhhar Siingh, Sugeeth Puranam, Vivek Gupta

FLARE: Faithful Logic-Aided Reasoning and Exploration (2025.emnlp-main)

Challenge:	a new benchmark is constructed to evaluate the accuracy of large language models for tabular data . the benchmark uses direct, indirect, and Chain-of-Thought prompting .
Approach:	They propose a framework that uses prompting, self-verification and constraint-based rule execution to improve accuracy.
Outcome:	The proposed framework significantly improves accuracy and recall in tabular data.

Copied to clipboard

Erik Arakelyan, Pasquale Minervini, Patrick Lewis, Pat Verga, Isabelle Augenstein

Sketch-of-Thought: Efficient LLM Reasoning with Adaptive Cognitive-Inspired Sketching (2025.emnlp-main)

Challenge:	Modern Question Answering (QA) and Reasoning approaches with Large Language Models (LLMs) use Chain-of-Thought (CoT) prompting but struggle with ambiguous tasks.
Approach:	They propose a method that uses large language models to plan solutions and formalize queries without external solvers to generate outputs faithful to their intermediate reasoning chains.
Outcome:	The proposed method achieves SOTA results on 7 out of 9 diverse reasoning benchmarks and 3 out of 3 logic inference benchmarks while enabling measurement of reasoning faithfulness.

Copied to clipboard

Simon A. Aytes, Jinheon Baek, Sung Ju Hwang

Evaluating Concurrent Robustness of Language Models Across Diverse Challenge Sets (2024.emnlp-main)

Challenge:	Recent advances in large language models (LLMs) have enabled strong reasoning capabilities through Chain-of-Thought (CoT) prompting.
Approach:	They propose a framework that integrates cognitively inspired reasoning paradigms with linguistic constraints to reduce token usage while preserving reasoning accuracy.
Outcome:	The proposed framework reduces token usage while preserving reasoning accuracy across 18 reasoning datasets across multiple domains, languages, and modalities.

Copied to clipboard

Vatsal Gupta, Pranshu Pandya, Tushar Kataria, Vivek Gupta, Dan Roth

Improving Large Language Models Function Calling and Interpretability via Guided-Structured Templates (2025.emnlp-main)

Challenge:	Language models display sensitivity to input perturbations, causing concerns about trust among users.
Approach:	They propose a methodology to examine how input perturbations affect language models across various scales, including pre-trained models and large language models.
Outcome:	The proposed methods enhance the model’s robustness to input perturbations and if exposure to one perturbation enhances or diminishes its performance with respect to other perturbations.

Copied to clipboard

Hy Dang, Tianyi Liu, Zhuofeng Wu, Jingfeng Yang, Haoming Jiang, Tao Yang, Pei Chen, Zhengyang Wang, Helen Wang, Huasheng Li, Bing Yin, Meng Jiang

Evaluating Fairness in Large Vision-Language Models Across Diverse Demographic Attributes and Prompts (2025.findings-emnlp)

Challenge:	Large language models (LLMs) have strong reasoning and tool-use capabilities, yet fail in real-world tool-interactions due to incorrect parameterization, poor tool selection, or misinterpretation of user intent.
Approach:	They propose a curriculum-inspired framework that leverages structured reasoning templates to guide LLMs through more deliberate step-by-step instructions for generating function calls.
Outcome:	The proposed framework reduces tool-use errors and improves interpretability and transparency of tool-using agents.

Copied to clipboard

Xuyang Wu, Yuan Wang, Hsin-Tai Wu, Zhiqiang Tao, Yi Fang

Looking Beyond the Pixels: Evaluating Visual Metaphor Understanding in VLMs (2025.findings-emnlp)

Challenge:	Large vision-language models have demonstrated strong capabilities in open-world visual understanding, but it is not clear how they address demographic biases in real life.
Approach:	They propose a method to assess visual fairness in LVLMs by question-answering/classification tasks.
Outcome:	The proposed approach improves transparency and offers a scalable solution for fairness mitigation.

Copied to clipboard

Manishit Kundu, Sumit Shekhar, Pushpak Bhattacharyya

Improving LLM-as-a-Judge Inference with the Judgment Distribution (2025.findings-emnlp)

Challenge:	Visual metaphors are a complex vision–language phenomenon that requires both perceptual and conceptual reasoning to understand.
Approach:	They introduce a visual metaphor dataset featuring 2177 synthetic and 350 human-annotated images and benchmark several SOTA VLMs on two tasks: Visual Metaphor Captioning (VMC) and Visual Metamorphosis VQA (VM-VQA).
Outcome:	The proposed model outperforms standard few-shot baselines on visual metaphors and VM-VQA tasks.

Copied to clipboard

Victor Wang, Michael JQ Zhang, Eunsol Choi

Challenge:	Using language models to scalably approximate human preferences on text quality (LLM-as-a-judge) is a standard practice applicable to many tasks.
Approach:	They propose to use LLM judges to approximate human preferences on text quality by using distributional output over judgment tokens.
Outcome:	The proposed method outperforms taking the mode (i.e. greedy decoding) in all evaluation settings, and incorporating risk aversion improves performance.

Code Execution as Grounded Supervision for LLM Reasoning (2025.emnlp-main)

Copied to clipboard

Dongwon Jung, Wenxuan Zhou, Muhao Chen

Entrospect: Information-Theoretic Self-Reflection Elicits Better Response Refinement of Small Language Models (2025.findings-acl)

Challenge:	Existing methods for generating high-quality CoT data rely on costly human annotations and error-prone CoT.
Approach:	They propose a method that extracts verifiable, step-by-step reasoning traces from code execution and transforms them into a natural language CoT reasoning.
Outcome:	The proposed method produces highly accurate reasoning data and reduces overall token length during inference by reducing meaningless repetition and overthinking.

Copied to clipboard

Tianqiang Yan, Ziqiao Lin, Lin Zhang, Zhenglong Sun, Yuan Gao

Token-Budget-Aware LLM Reasoning (2025.findings-acl)

Challenge:	Existing approaches to self-reflection fail to deliver robust response refinement for models with parameter sizes of 10 billion or smaller.
Approach:	They propose to redesign Self-Refine and introduce an information-theoretic framework based on Chain-of-Thought prompt engineering to improve self-reflection in Small Language Models.
Outcome:	The proposed framework improves reasoning accuracy and computational efficiency by up to 36.2% under identical model and data settings.

Copied to clipboard

Tingxu Han, Zhenting Wang, Chunrong Fang, Shiyu Zhao, Shiqing Ma, Zhenyu Chen

Self-Training Elicits Concise Reasoning in Large Language Models (2025.findings-acl)

Challenge:	Existing methods to enhance reasoning capabilities of large language models incur significant overhead in token usage, leading to increased costs.
Approach:	They propose a token-budget-aware LLM reasoning framework that adjusts the number of reasoning tokens based on the reasoning complexity of each problem.
Outcome:	The proposed method reduces token costs in CoT reasoning with only a slight performance reduction.

Copied to clipboard

Tergel Munkhbat, Namgyu Ho, Seo Hyun Kim, Yongjin Yang, Yujin Kim, Se-Young Yun

Generative Reward Modeling via Synthetic Criteria Preference Learning (2025.acl-long)

Challenge:	Chain-of-thought reasoning has enabled large language models to use additional computation through intermediate tokens to solve complex tasks, but current models often generate more tokens than necessary to accomplish the task, incurring extraneous inference costs.
Approach:	They propose to fine-tune models with self-generated concise reasoning paths obtained by best-of-N sampling and few-shot conditioning in task-specific settings to elicit concise reasoning.
Outcome:	The proposed method reduces output tokens by 30% on GSM8K and MATH while maintaining average accuracy.

Copied to clipboard

Xiaobo Liang, Haoke Zhang, Juntao Li, Kehai Chen, Qiaoming Zhu, Min Zhang

Small Models Struggle to Learn from Strong Reasoners (2025.findings-acl)

Challenge:	Generative Reward Models (GenRMs) leverage synthesized Chains of Thought (CoT) but this approach introduces risks of overoptimization due to the inability to guarantee the correctness of the CoTs.
Approach:	They propose a criteria-based preference tree for GenRMs that uses chain of thought to generate reasoning . they show that synthesized data can be learned using a long CoT format .
Outcome:	The proposed model shows significant improvements over baselines on multiple human preference benchmarks.

Copied to clipboard

Yuetai Li, Xiang Yue, Zhangchen Xu, Fengqing Jiang, Luyao Niu, Bill Yuchen Lin, Bhaskar Ramasubramanian, Radha Poovendran

Exploring Chain-of-Thought Reasoning for Steerable Pluralistic Alignment (2025.emnlp-main)

Challenge:	a small learning gap exists between large and small language models . long CoT data and large model responses are not beneficial for small models - a problem that may be due to the small student model's ability to handle distribution shifts.
Approach:	They propose a mix distillation strategy that balances reasoning complexity by combining long and short CoT examples or reasoning from both larger and smaller models.
Outcome:	The proposed strategy outperforms training on large and small models on short CoT and small model CoT.

Copied to clipboard

Yunfan Zhang, Kathleen McKeown, Smaranda Muresan

Logical Reasoning with Outcome Reward Models for Test-Time Scaling (2025.emnlp-main)

Challenge:	Large Language Models (LLMs) are typically trained to reflect a relatively uniform set of values, which limits their applicability to tasks that require understanding of nuanced human perspectives.
Approach:	They propose to use Chain-of-Thought reasoning techniques to build steerable pluralistic models by fine-tuning on human-authored CoT and synthetic explanations.
Outcome:	The proposed methods outperform others and demonstrate strong sample efficiency.

Copied to clipboard

Ramya Keerthy Thatikonda, Wray Buntine, Ehsan Shareghi

Think Faster Than Words: Efficient LLM Chain-of-Thought Reasoning via Dynamic Shortcut Decoding (2026.acl-long)

Challenge:	Logical reasoning is a critical benchmark for evaluating the capabilities of large language models (LLMs), but it is under-explored in deductive reasoning.
Approach:	They propose to use Chain-of-Thought to generate data using single and multiple samples to train ORMs.
Outcome:	The proposed model expands the type of errors covered in the training dataset, covering previously unexplored error types.

Copied to clipboard

Fan Liu, Yanhao Wang, Min Zhang, Zhikang Chen, Zeyuan Li, Lewei He, Jiahui Pan

InspireDebate: Multi-Dimensional Subjective-Objective Evaluation-Guided Reasoning and Optimization for Debating (2025.acl-long)

Challenge:	Existing methods that prune or employ early stopping to reduce latency often compromise reasoning reliability.
Approach:	They propose a shortcut decoding framework that integrates probes over internal hidden states with step-level entropy to detect convergence of reasoning during generation and adaptively selects between a fast-exit path and a stability-verified path to remove redundant steps while preserving answer correctness.
Outcome:	The proposed framework reduces token usage by approximately 35% and maintains accuracy comparable to full CoT decoding.

Copied to clipboard

Fuyu Wang, Jiangtong Li, Kun Zhu, Changjun Jiang

ReasonMed: A 370K Multi-Agent Generated Dataset for Advancing Medical Reasoning (2025.emnlp-main)

Challenge:	Existing LLMs focus on responding to specific arguments while neglecting objective assessments such as authenticity and logical validity.
Approach:	They propose a multi-dimensional evaluation system and an optimized debating framework . they propose to use coT reasoning enhancement, web-based Retrieval Augmented Generation to optimize across various dimensions.
Outcome:	The proposed framework outperforms baseline models in argument quality assessment and debate process simulation by 57%.

Copied to clipboard

Yu Sun, Xingyu Qian, Weiwen Xu, Hao Zhang, Chenghao Xiao, Long Li, Deli Zhao, Wenbing Huang, Tingyang Xu, Qifeng Bai, Yu Rong

Table-R1: Region-based Reinforcement Learning for Table Understanding (2026.findings-acl)

Challenge:	Existing medical reasoning datasets are limited in scale and typically rely on incomplete data.
Approach:	They propose to use ReasonMed to train medical reasoning models using a multi-agent generation, verification, and refinement pipeline.
Outcome:	The largest medical reasoning dataset to date surpasses the prior best sub-10B models by 4.17% and even exceeds LLaMA3.1-70B on PubMedQA by 4.60%.

Copied to clipboard

Zhenhe Wu, Jian Yang, Zhongjiang He, Changzai Pan, Jiaheng Liu, Xianjie Wu, Yu Zhao, Shuangyong Song, Yongxiang Li, Zhoujun Li, Xuelong Li

Learning from Mistakes: Negative Reasoning Samples Enhance Out-of-Domain Generalization (2026.acl-long)

Challenge:	Tables are a widely used data format that poses unique challenges for language models due to their structured row-column interactions.
Approach:	They propose a region-based reinforcement learning approach that integrates region evidence into reasoning steps.
Outcome:	The proposed method outperforms baseline models on three benchmark datasets and significantly reduces the reasoning token consumption by 67.5%.

Copied to clipboard

Tian Xueyun, MingHua Ma, Bingbing Xu, Nuoyan Lyu, Wei Li, Heng Dong, Zheng Chu, Yuanzhuo Wang, Huawei Shen

Think Better, Not Longer: Token-Level Marginal Utility for Efficient Reasoning in Large Reasoning Models (2026.acl-long)

Challenge:	Recent studies show that supervised fine-tuning (SFT) is a common approach for reasoning in large language models.
Approach:	They propose to use supervised fine-tuning (SFT) on chain-of-thought trajectories demonstrations . they find that incorporating negative traxories yields substantial OOD generalization gains .
Outcome:	The proposed scheme yields 5.51% OOD gain over positive-only training.

Copied to clipboard

Jiawei Li, Yang Gao, Huashan Sun, Chong Feng

To Judge or Not to Judge: Can Large Language Models Leverage the Dispute Focus in Legal Judgment? (2026.acl-long)

Challenge:	Large reasoning models (LRMs) generate explicit Chain-of-Thought rationales, but often suffer from "overthinking".
Approach:	They propose a unified training framework to synthesize concise reasoning chains by identifying tokens that reduce the model’s likelihood of the correct answer.
Outcome:	Experiments on deepSeek-R1-Distill-Qwen backbones show that MUTO yields better efficiency-accuracy Pareto frontier.

Copied to clipboard

Luoming Hu, Liang Yang, Jingjie Zeng, Zijie Xing

UniVocal: Unified Speech-Singing Code-Switching Synthesis (2026.acl-long)

Challenge:	Existing research on large language models for legal judgment prediction fails to address the complexity of civil judicial cases.
Approach:	They propose a framework that leverages the dispute focus to guide LLMs through a structured, judge-like cognitive workflow.
Outcome:	The proposed framework can guide LLMs through a structured, judge-like cognitive workflow.

Copied to clipboard

YuFei Shi, Qian Chen, Wen Wang, Xiangang Li, Zhen-Hua Ling, Yang Ai

Speech & Audio in NLP

Challenge:	Existing systems cannot automatically determine when to switch between modes based on text content.
Approach:	They propose a unified framework that implicitly infers vocal modes from text context to pioneer SCS Synthesis.
Outcome:	The proposed framework infers vocal modes solely from text context to pioneer SCS Synthesis.

Trustworthiness and Self-awareness in Large Language Models: An Exploration through the Think-Solve-Verify Framework (2024.lrec-main)

Copied to clipboard

Zhendong Liu, Changhong Xia, Wei He, Chongjun Wang

Self-SoftCoT: A Self-Consistent Framework via Position-Aware Latent Space Reinforcement Learning (2026.acl-long)

Challenge:	Large Language Models (LLMs) are becoming increasingly influential in reasoning tasks, but they lack trustworthiness and introspective self-awareness when subjected to complex reasoning tasks.
Approach:	They propose a framework to explore LLMs’ trustworthiness, introspective self-awareness, and collaborative reasoning by using the Think-Solve-Verify framework.
Outcome:	The proposed approach improves from 67.3% to 72.8% on the AQuA dataset and demonstrates the model’s ability to explain the given answers.

Copied to clipboard

Liangliang Dong, Lianlei Shan, Shuaimin Li

Analysing Chain of Thought Dynamics: Active Guidance or Unfaithful Post-hoc Rationalisation? (2025.emnlp-main)

Challenge:	Existing Continuous reasoning approaches rely on external auxiliary models, resulting in complex deployment and fractured inference pipelines.
Approach:	They propose a self-contained framework that enables a frozen LLM to internally generate and consume latent thoughts without external assistants.
Outcome:	The proposed framework outperforms SoftCoT models on five reasoning benchmarks.

Copied to clipboard

Samuel Lewis-Lim, Xingwei Tan, Zhixue Zhao, Nikolaos Aletras

LoRE: Enhancing Search Relevance with Progressive Chain-of-Thought and Preference Alignment (2026.findings-acl)

Challenge:	Recent work has demonstrated that using chain of thought (CoT) on soft-reasoning tasks can yield limited or even negative performance gains.
Approach:	They investigate how chain of thought (CoT) is used in soft-reasoning tasks across instruction-tuned, reasoning and reasoning-distilled models.
Outcome:	The proposed model can steer predictions without faithfully reflecting reasoning, indicating a disconnect between CoT influence and faithfulness.

Copied to clipboard

Chenji Lu, Zhuo Chen, Hui Zhao, Zhiyuan Zeng, Gang Zhao, Junjie Ren, null Lihaoran, Songyan Liu, Pengjie Wang, Chuan Yu, Jian Xu, Bo Zheng

VET: Verifiable Execution Tracing for Reliable Text-to-SQL Generation (2026.findings-acl)

Challenge:	E-commerce search relevance is a critical component of retrieval systems.
Approach:	They propose a large-generative model for search relevance that trains reasoning knowledge, multi-modal understanding and rule awareness into three core competencies.
Outcome:	The proposed model outperforms GPT-5 in Macro-F1 and achieves 27% online gain.

Copied to clipboard

Dongyu Wang, Jingyu Li, Lan Zhang, null Ganggang.yu, Liang Huang

Training LLMs for Divide-and-Conquer Reasoning Elevates Test-Time Scalability (2026.acl-long)

Challenge:	Existing methods for text-to-SQL generation are prone to hallucinations and grounding . authors present a novel reasoning paradigm that transforms text- to-Sql from unverifiable textual rationales into step-wise executable semantics.
Approach:	They propose a reasoning paradigm that transforms text-to-SQL from unverifiable textual rationales into step-wise executable semantics.
Outcome:	The proposed reasoning paradigm transforms text-to-SQL from unverifiable textual rationales into step-wise executable semantics.

Copied to clipboard

Xiao Liang, Zhong-Zhi Li, Zhenghao Lin, Eric Hanchen Jiang, Hengyuan Zhang, Yelong Shen, Kai-Wei Chang, Ying Nian Wu, Yeyun Gong, Weizhu Chen

CityVG: Contrastive Fine-Tuning and Reward-Based Chain-of-Thought Reasoning for Zero-Shot City-Scale 3D Visual Grounding (2026.acl-long)

Challenge:	Large language models have demonstrated strong reasoning capabilities through step-by-step chain-of-thought (CoT) reasoning, but their strictly sequential nature constrains test-time scalability.
Approach:	They propose an end-to-end reinforcement learning framework to enhance LLMs' DAC-style reasoning capacity by decomposing a problem into subproblems and solving them sequentially.
Outcome:	The proposed model surpasses CoT by 8.6% and 6.3% on competition-level benchmarks and is available at the [github.com/MasterVito/DAC-RL].

Copied to clipboard

Jianjun Zhang, Hanli Wang

PICTURE: Enhancing Theory-of-Mind in Large Language Models by Revealing, Not Hiding, Characters’ Lack of Knowledge (2026.acl-long)

Challenge:	Existing methods for 3D visual grounding are limited to small-scale indoor data or require heavy supervision.
Approach:	They propose a contrastive fine-tuning strategy to align textual queries with urban scene graphs.
Outcome:	The proposed framework achieves strong zero-shot localization performance and generalizes effectively to unseen urban environments.

Copied to clipboard

Eojin Jeon, SangKeun Lee

Interpretable Traces, Unexpected Outcomes: Investigating the Disconnect in Trace-Based Knowledge Distillation (2026.acl-long)

Challenge:	Existing approaches to simulating Theory of Mind (ToM) using LLMs have been a longstanding problem in natural language processing (NLP).
Approach:	They propose a new method that enables LLMs to generate a character’s lack of knowledge within free-form Chain-of-Thought (CoT) based on this method, they propose to generate perspective-taking outputs as free- form explanations without event hiding.
Outcome:	The proposed method outperforms existing prompting methods by an average of 7.3% on false-belief tasks.

Copied to clipboard

Siddhant Bhambri, Upasana Biswas, Subbarao Kambhampati

Beyond Markovian Forgetfulness: Episodic Memory for Reasoning-Intensive Retrieval (2026.acl-long)

Challenge:	Recent advances in reasoning-oriented Large Language Models have been driven by the introduction of Chain-of-Thought (CoT) traces.
Approach:	They propose to use CoT traces to guide model inference and serve as supervision signals for Knowledge Distillation to improve smaller models.
Outcome:	The proposed model is based on a rule-based problem decomposition method and is valid for both semantic correctness and interpretability to the end user.

Copied to clipboard

Dohyeon Lee, Yeonseok Jeong, Seung-won Hwang

Turning Logic Against Itself: Probing Model Defenses Through Contrastive Questions (2025.emnlp-main)

Challenge:	Existing methods for reasoning-intensive information retrieval suffer from inefficiency . Chain-of-Thought (CoT) approaches suffer from lack of token efficiency . Existing models lack episodic memory, which stores the history of prior states .
Approach:	They propose an algorithm that enhances state-based frameworks with an episodic memory module that stores the full history of prior states for a query.
Outcome:	The proposed model outperforms CoT and state-based models on the BRIGHT benchmark and is highly token-efficient.

Copied to clipboard

Rachneet Singh Sachdeva, Rima Hazra, Iryna Gurevych

CAMEC: Complexity-Aware Multi-Expert Collaboration for Reliable Chinese Medical Question Answering (2026.acl-long)

Challenge:	Existing safety measures detect overt malicious intent but fail to address subtle, reasoning-driven vulnerabilities.
Approach:	They propose a two-phase jailbreak technique that exploits contrastive reasoning to bypass safety mechanisms in large language models.
Outcome:	The proposed techniques achieve higher attack success rates (44%) than existing methods.

Copied to clipboard

Yukang Wu, Xiyuan Jia, Jiayi Wu, HongChen Yu, Yuhan Qiu, Wu Guohua

Question Answering

Challenge:	Large language models are promising for medical question answering in china, but remain unreliable due to hallucinations, weak factual grounding and difficulty handling clinically complex cases.
Approach:	They propose a framework that combines hierarchical medical adaptation with complexity-aware expert routing for reliable Chinese medical QA.
Outcome:	The proposed framework outperforms strong general and medical LLM baselines on four Chinese medical benchmarks.

Too Long, Do Re-weighting for Efficient LLM Reasoning Compression (2026.acl-long)

Copied to clipboard

Zhong-Zhi Li, Xiao Liang, Zihao Tang, Lei Ji, Peijie Wang, Haotian Xu, Xing W, Haizhen Huang, Weiwei Deng, Yeyun Gong, Zhijiang Guo, Xiao Liu, Fei Yin, Cheng-Lin Liu

Evaluating Large Vision Language Models on Bangla Medical Visual Question Answering (2026.findings-acl)

Challenge:	Large Language Models (LLMs) have recently achieved remarkable progress on complex reasoning tasks by leveraging extended Chain-of-Thought (CoT) techniques.
Approach:	They propose a method that uses Extended Chain-of-Thought (EFT) to reduce the number of output tokens by nearly 40% while maintaining the accuracy of the reasoning.
Outcome:	The proposed method reduces the number of output tokens by nearly 40% while maintaining the accuracy of the reasoning.

Copied to clipboard

Rafid Ahmed, Intesar Tahmid, Mir Sazzat Hossain, Tasnimul Hossain Tomal, Md Mahir Jawad, Anam Borhan Uddin, Md Fahim, Md Farhad Alam Bhuiyan

AdaMix: Adaptive Mixing for Short and Long Reasoning Adapters (2026.acl-long)

Challenge:	Recent advances in Large Language Models and Large Vision Language Model (LVLMs) have demonstrated promising capabilities in complex reasoning tasks, but low-resource contexts like Bangla are underexplored.
Approach:	They propose a multilingual medical visual question answering dataset using Bangla.
Outcome:	The proposed model performs well on generalized visual tasks but struggles with fine-grained diagnostic reasoning, achieving low accuracy in specialized categories.

Copied to clipboard

Hao Luo, Xiao Yan, Xinyan Li, Qiming Zeng, Yuhao Lin, Shanshan Feng, Hao Wang, Jiawei Jiang

CoTEvol: Self-Evolving Chain-of-Thoughts for Data Synthesis in Mathematical Reasoning (2026.findings-acl)

Challenge:	Existing methods for large reasoning models have improved efficiency but still face limitations such as conflicting objectives and limited adaptability.
Approach:	They propose an adaptive reasoning framework that applies a uniform, computation-intensive deep reasoning strategy to all problems.
Outcome:	The proposed framework reduces the average response length of DeepSeek-R1-Distill-Qwen-7B by 54.9% while improving accuracy by up to 4.8% on five mathematical datasets.

Copied to clipboard

Zhuo Wang, Zhuo Zhang, Yafu Li, Yu Cheng, Lizhen Qu, Zenglin Xu

Time-for-Accuracy: Formalizing Chain-of-Thought as an Expansion of Logical Depth (2026.findings-acl)

Challenge:	Existing methods to improve LLMs' reasoning abilities suffer from diminishing returns or high computing overhead.
Approach:	They propose a genetic evolutionary framework that casts CoT generation as a population-based search over reasoning trajectories.
Outcome:	The proposed framework improves correct-CoT synthesis success by over 30% and enhances structural diversity with markedly improved efficiency.

Copied to clipboard

Yue Wang, Zhi Zhang, Wang Xi, Chengjie Sun, Lili Shan, Bingquan Liu

Challenge:	Chain-of-thought (CoT) prompting can improve multi-step reasoning, but it is unclear what kind of additional sequential computation longer traces actually enable.
Approach:	They propose a deletion-based measure of step necessity under a specified inference interface to operationalize realized depth beyond raw length.
Outcome:	The proposed method combines effective logical depth with Bennett's logical depth to show that it is more efficient than a linear model.

Valid Necessary: Diagnosing Latent Inefficiency in Chain-of-Thought (2026.findings-acl)

Copied to clipboard

Daeyeop Lee, Hwanjo Yu

CRISP: Compressing Redundancy in Chain-of-Thought via Intrinsic Saliency Pruning (2026.findings-acl)

Challenge:	Existing reasoning step evaluators fail to distinguish “valid but inefficient” reasoning steps from necessary reasoning.
Approach:	They propose a training-free metric that identifies low-utility steps and a post-hoc compression strategy to quantify their impact on token usage.
Outcome:	The proposed metric reduces token consumption by 31–53% while maintaining accuracy at substantially higher compression rates.

Copied to clipboard

Yangsong Lan, Hongliang Dai, Piji Li

Thinking Economically: A Hierarchical Framework for Adaptive-Complexity Reasoning in LLMs (2026.findings-acl)

Challenge:	Existing approaches to compress CoT with external compressors fail to align with the model’s internal reasoning dynamics, resulting in the loss of critical logical steps.
Approach:	They propose a framework that exploits the model’s intrinsic saliency to compress CoT by exploiting its reasoning termination token .
Outcome:	The proposed framework reduces redundancy in reasoning chain by exploiting the model’s intrinsic saliency.

Copied to clipboard

Yubo Gao, Haotian Wu, Hong Chen, Junquan Huang, Yibo Yan, Jungang Li, Zihao Dongfang, Sicheng Tao, PS Tan, Jie Zhang, Xuming Hu

MIND: From Passive Mimicry to Active Reasoning through Capability-Aware Multi-Perspective CoT Distillation (2026.acl-long)

Challenge:	Existing efficiency methods for Chain-of-Thought (CoT) generate excessively long rationales without commensurate accuracy gains.
Approach:	They propose a training framework that operationalizes this principle through coarse-to-fine budgeting.
Outcome:	Experiments on GSM8K and MATH500 show that HAB surpasses standard CoT in accuracy and reduces token usage, achieving stronger performance-efficiency trade-off than baselines.

Copied to clipboard

Jin Cui, Jiaqi Guo, Jiepeng Zhou, Ruixuan Yang, Jiayi Lu, Jiajun Xu, Jiangcheng Song, Boran Zhao, Pengju Ren

rSIM: Incentivizing Reasoning Capabilities of LLMs via Reinforced Strategy Injection (2026.acl-long)

Challenge:	Existing approaches restrict students to following a single golden rationale and treat different reasoning paths independently, causing suboptimal performance.
Approach:	They propose a capability-adaptive framework that transitions distillation from passive mimicry to active cognitive construction and employ a feedback-driven inertia calibration mechanism to align supervision with the student’s current adaptability.
Outcome:	Experiments show that the proposed framework achieves state-of-the-art performance on both in-distribution and out-of distribution benchmarks.

Copied to clipboard

Sijia Chen, Baochun Li, Di Niu

Choose Your Lens: Multi-Perspective Value Alignment of Chain-of-Thought Reasoning (2026.findings-acl)

Challenge:	Existing literature on Reasoning Language Models (RLMs) focuses on the ability to integrate reasoning strategies into the chain-of-thought process, contributing to improved problem-solving accuracy.
Approach:	They propose a reinforced strategy injection mechanism that enables any LLM to become an RLM by employing a small planner to guide the LLM's CoT through the adaptive injection of reasoning strategies.
Outcome:	The proposed model outperforms existing models in mathematical, coding, and financial reasoning tasks and is generalizable.

Copied to clipboard

Gejian Zhao, Hanzhou Wu, Xinpeng Zhang

Scaling Evaluation-Time Compute with Reasoning Models as Evaluators (2026.findings-acl)

Challenge:	Large language models tend to hallucinate “convenient” facts to forcefully justify stances . current methods often induce motivated reasoning, causing factual hallucinations .
Approach:	They propose a neuro-symbolic framework that enables steerable pluralism without distorting objective reality by projecting generated CoT paths onto a multi-perspective graph.
Outcome:	The proposed approach reduces factual hallucinations by 3 and improves cross-perspective consistency by 25% compared to standard steerable baselines, paving the way for trustworthy pluralistic AI.

Copied to clipboard

Seungone Kim, Ian Wu, Jinu Lee, Xiang Yue, Seongyun Lee, Minkyeong Moon, Carolin Lawrence, Kiril Gashteovski, Julia Hockenmaier, Graham Neubig, Sean Welleck

Is Chain-of-Thought Really Not Explainability? Chain-of-Thought Can Be Faithful without Hint Verbalization (2026.acl-long)

Challenge:	Language model (LM) evaluators that generate chain-of-thought reasoning are widely used for the assessment of LM responses.
Approach:	They investigate whether increasing LMs' "thinking" time through scaling test-time compute can improve an LM's evaluation capability.
Outcome:	The proposed reasoning models improve evaluation performance monotonically with the number of reasoning tokens generated, mirroring trends seen in LM reasoning.

Copied to clipboard

Kerem Zaman, Shashank Srivastava