Papers with reasoning
Structure Learning for Neural Module Networks (D19-64)
Copied to clipboard
| Challenge: | Neural Module Networks are a class of neural networks that involve human-specified neural modules . current models only learn the parameters of the modules and/or the order of their execution . |
| Approach: | They propose to learn internal structure and sequence without extra supervisory signals . they use dynamically composable modules which are then assembled into a layout . |
| Outcome: | The proposed model performs comparable to models using hand-designed modules. |
From Multimodal LLM to Human-level AI: Modality, Instruction, Reasoning, Efficiency and beyond (2024.lrec-tutorials)
Copied to clipboard
| Challenge: | This tutorial aims to deliver a comprehensive review of cutting-edge research in MLLMs. |
| Approach: | This tutorial will review cutting-edge research in MLLMs and examine the impact of ML in learning and reasoning. |
| Outcome: | This course will review cutting-edge research in MLLMs and examine the impact of ML models on learning, learning, and multimodal reasoning. |
Complex Reasoning in Natural Language (2023.acl-tutorials)
Copied to clipboard
| Challenge: | Recent research shows that pretrained language models are often brittle for complex reasoning tasks. |
| Approach: | They propose to use pre-trained language models to teach machines to reason over texts . they will review recent promising approaches to tackling complex reasoning tasks . |
| Outcome: | This tutorial reviews promising approaches to complex reasoning tasks . it reviews the methods that can be used to augment models with robustness . |
Deep Learning for Natural Language Inference (N19-5)
Copied to clipboard
| Challenge: | This tutorial discusses cutting-edge research on NLI, including recent advance on dataset development, cutting- edge deep learning models, and highlights from recent research on using NLI to understand capabilities and limits of deep learning for language understanding and reasoning. |
| Approach: | This tutorial discusses cutting-edge research on NLI, including recent advance on dataset development and cutting- edge deep learning models. |
| Outcome: | This tutorial discusses cutting-edge research on NLI, including recent advance on dataset development, cutting- edge deep learning models, and highlights from recent research on using NLI to understand capabilities and limits of deep learning model for language understanding and reasoning. |
RAVEN: Robust Advertisement Video Violation Temporal Grounding via Reinforcement Reasoning (2025.acl-industry)
Copied to clipboard
| Challenge: | Existing methods for detecting ads video violations lack precise temporal grounding, noisy annotations, and limited generalization. |
| Approach: | They propose a framework that integrates curriculum reinforcement learning with large language models to enhance reasoning and cognitive capabilities for violation detection. |
| Outcome: | The proposed framework achieves superior performance in violation category accuracy and temporal interval localization. |
Being Kind Isn’t Always Being Safe: Diagnosing Affective Hallucination in LLMs (2026.findings-eacl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are increasingly engaged in emotionally vulnerable conversations that extend beyond information seeking to moments of personal distress. |
| Approach: | They propose AHaBench, a benchmark of 500 mental-health-related prompts with expert-informed reference responses, evaluated along three dimensions: Emotional Enmeshment, Illusion of Presence, and Fostering Overdependence. |
| Outcome: | The proposed model is based on 500 mental-health-related prompts with expert-informed reference responses and a 5K-instance preference dataset enabling direct preference optimization (DPO) for alignment with emotionally responsible behavior. |
Why Large Language Models can Secretly Outperform Embedding Similarity in Information Retrieval (2026.acl-srw)
Copied to clipboard
| Challenge: | Recent studies show that similarity is a short-sighted interpretation of relevance . LLM-based Relevance Judgment Systems (LLM-RJS) can overcome this limitation . |
| Approach: | They propose that LLM-Based Relevance Judgment Systems can overcome short-sighted interpretation of relevance by embedding similarity instead of similarity. |
| Outcome: | The proposed methods outperform Neural Embedding Retrieval Systems by overcoming similarity limitation. |
Representation, Learning and Reasoning on Spatial Language for Downstream NLP Tasks (2020.emnlp-tutorials)
Copied to clipboard
| Challenge: | In this tutorial, we discuss the cutting-edge research results and existing challenges related to spatial language understanding including semantic annotations, existing corpora, symbolic and sub-symbolic representations, qualitative spatial reasoning, spatial common sense, deep and structured learning models. |
| Approach: | This tutorial presents cutting-edge research results and current challenges related to spatial language understanding including semantic annotations, existing corpora, symbolic and sub-symbolic representations, qualitative spatial reasoning, spatial common sense, deep and structured learning models. |
| Outcome: | This paper reviews the cutting-edge research results and current challenges related to spatial language understanding including semantic annotations, existing corpora, symbolic and sub-symbolic representations, qualitative spatial reasoning, spatial common sense, deep and structured learning models. |
Archer: A Human-Labeled Text-to-SQL Dataset with Arithmetic, Commonsense and Hypothetical Reasoning (2024.eacl-long)
Copied to clipboard
| Challenge: | Existing text-to-SQL datasets that capture complex reasoning are limited by their execution accuracy. |
| Approach: | They present a bilingual text-to-SQL dataset specific to complex reasoning . their evaluation shows that Archer challenges the capabilities of current models . |
| Outcome: | The proposed dataset challenges state-of-the-art models with 6.73% execution accuracy . the dataset contains 1,042 English and 1,042, Chinese questions and 521 unique SQL queries . |
Collecting Diverse Natural Language Inference Problems for Sentence Representation Evaluation (D18-1)
Copied to clipboard
Adam Poliak, Aparajita Haldar, Rachel Rudinger, J. Edward Hu, Ellie Pavlick, Aaron Steven White, Benjamin Van Durme
| Challenge: | a plethora of new natural language inference datasets has been created in recent years . however, these datasets do not provide clear insight into what type of reasoning or inference a model may be performing. |
| Approach: | They propose to recast 13 existing natural language inference datasets into a common structure. |
| Outcome: | The proposed datasets provide insight into how well a sentence representation captures distinct types of reasoning. |
Deep Reinforcement Learning for NLP (P18-5)
Copied to clipboard
| Challenge: | Many natural language processing tasks can be formulated as deep reinforcement learning (DRL) problems. |
| Approach: | This tutorial provides an introduction to the foundations of deep reinforcement learning . it describes recent advances in designing deep reinforcement for NLP . |
| Outcome: | This tutorial provides an introduction to the foundations of deep reinforcement learning and some practical solutions for NLP tasks. |
From Factuality to Meta-Factivity: A Cognitive Blueprint for Trustworthy LLMs (2026.acl-short)
Copied to clipboard
| Challenge: | Current evaluation paradigms on Event Factuality Prediction (EFP) focus on static classification tasks and shortcut learning and unreliable reasoning. |
| Approach: | They propose a meta-factivity framework that moves evaluation beyond surface recognition to belief trajectory reasoning and epistemic regulation. |
| Outcome: | The proposed framework shifts from event factuality to meta-factivity . the proposed framework lays the groundwork for a more rigorous benchmark for explainable self-governance . |
Book QA: Stories of Challenges and Opportunities (D19-58)
Copied to clipboard
| Challenge: | Existing approaches to answer questions based on the full text of books are limited by their unique characteristics. |
| Approach: | They propose a system for answering questions based on the full text of books . they use a memory network to reason and predict an answer, and a novel question generator to improve generalization. |
| Outcome: | The proposed system improves on the recently published NarrativeQA corpus on Who questions . it shows that the proposed system is highly challenging and needs more research . |
IDP Accelerator: Agentic Document Intelligence from Extraction to Compliance Validation (2026.acl-demo)
Copied to clipboard
Md Mofijul Islam, Md Sirajus Salekin, Joe King, Priyashree Roy, Vamsi Thilak Gudi, Spencer Romo, Akhil Nooney, Bob Strahan, Boyi Xie, Diego A. Socolinsky
| Challenge: | Large Language Models (LLMs) are inadequate for extracting structured insights from unstructured documents. |
| Approach: | They propose a framework enabling agentic AI for end-to-end document intelligence with four key components: DocSplit, configurable Extraction Module, and Rule Validation Module. |
| Outcome: | The proposed framework achieves 98% classification accuracy, 80% reduced processing latency, and 77% lower operational costs over legacy baselines. |
Unifying Language Agent Algorithms with Graph-based Orchestration Engine for Reproducible Agent Research (2025.acl-demo)
Copied to clipboard
Qianqian Zhang, Jiajia Liao, Heting Ying, Yibo Ma, Haozhan Shen, Jingcheng Li, Peng Liu, Lu Zhang, Chunxin Fang, Kyusong Lee, Ruochen Xu, Tiancheng Zhao
| Challenge: | Language agents powered by large language models (LLMs) have demonstrated remarkable capabilities in understanding, reasoning, and executing complex tasks. |
| Approach: | They propose a flexible framework that addresses engineering overhead and insufficient evaluation frameworks for fair comparison. |
| Outcome: | The proposed framework simplifies language agent development and establishes a foundation for reproducible agent research. |
Sketching a Linguistically-Driven Reasoning Dialog Model for Social Talk (2022.acl-srw)
Copied to clipboard
| Challenge: | a new study shows that dialog systems that can hold social talk and make sense of conversational content are not efficient for context-sensitive natural language understanding and reasoning. |
| Approach: | They propose a linguistically-informed architecture to handle social talk in English . they propose linguistic models that fit the context-sensitive components into a Bayesian game-theoretic model . |
| Outcome: | The proposed architecture is based on corpus-based methods but does not track what is happening in a conversation. |
Calibrating LLM Judges: Linear Probes for Fast and Reliable Uncertainty Estimation (2026.acl-industry)
Copied to clipboard
Bhaktipriya Radharapu, Eshika Saxena, Kenneth Li, Chenxi Whitehouse, Adina Williams, Nicola Cancedda
| Challenge: | Existing methods for obtaining well-calibrated uncertainty estimates are poorly calibrated or computationally expensive. |
| Approach: | They propose a linear probe that provides calibrated uncertainty estimates from reasoning judges’ hidden states, requiring no additional model training. |
| Outcome: | The proposed method achieves superior calibration compared to existing methods with x computational savings, generalizes robustly to unseen evaluation domains, and delivers higher accuracy on high-confidence predictions. |
If Probable, Then Acceptable? Understanding Conditional Acceptability Judgments in Large Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Conditional acceptability refers to how plausible a conditional statement is perceived to be. |
| Approach: | They propose to use conditional acceptability models to assess their models' conditional acceptance . they use probabilistic and semantic cues to assess whether a condition is acceptable or plausible . |
| Outcome: | The proposed models are sensitive to conditional probability and semantic relevance, but do not align more closely with human judgments. |
UnifiedMLLM: Enabling Unified Representation for Multi-modal Multi-tasks With Large Language Model (2025.findings-naacl)
Copied to clipboard
Zhaowei Li, Wei Wang, YiQing Cai, Qi Xu, Pengyu Wang, Dong Zhang, Hang Song, Botian Jiang, Zhida Huang, Tao Wang
| Challenge: | Representative models like LLaVA and MiniGPT-4 have great capabilities in various tasks. |
| Approach: | They propose a unified model to represent various multi-modal tasks using a single representation. |
| Outcome: | The proposed model outperforms existing models in a variety of tasks while maintaining generality and scalability. |
Capture the Key in Reasoning to Enhance CoT Distillation Generalization (2025.acl-long)
Copied to clipboard
| Challenge: | Existing distillation methods for Large Language Models (LLMs) focus on fine-tuning student SLMs on correct data, resulting in students struggling to learn the key instead of analyzing mistakes according to correct solutions. |
| Approach: | They propose a method that exposes key reasoning steps rather than simple fine-tuning students' CoTs data by using a set of prompts with similar reasoning paths but divergent conclusions. |
| Outcome: | The proposed method improves student SLMs' ability to learn key reasoning steps rather than fine-tuning them on teacher data. |
EXPLORER: Exploration-guided Reasoning for Textual Reinforcement Learning (2024.eacl-long)
Copied to clipboard
Kinjal Basu, Keerthiram Murugesan, Subhajit Chaudhury, Murray Campbell, Kartik Talamadupula, Tim Klinger
| Challenge: | Text-based games (TBGs) combine natural language understanding with reasoning. |
| Approach: | They propose an exploration-guided reasoning agent for textual reinforcement learning that integrates natural language with reasoning. |
| Outcome: | The proposed agent outperforms baseline agents on TWG and TWC games. |
Verifying the Steps of Deductive Reasoning Chains (2025.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models have been shown to improve the reasoning capabilities of the models. |
| Approach: | They propose to automate verification of individual reasoning steps in a logical deductive Chain-of-Thought. |
| Outcome: | The proposed method can detect unsound reasoning steps fairly well, but under-performs symbolic methods. |
Relation Embedding with Dihedral Group in Knowledge Graph (P19-1)
Copied to clipboard
| Challenge: | Existing models for link prediction fail to capture relation compositions, resulting in lack of interpretability for reasoning on incomplete knowledge graph (KG). |
| Approach: | They propose a new model that learns knowledge graph embeddings that can capture relation compositions by nature and reduces the solution space drastically. |
| Outcome: | The proposed model outperforms existing models and is comparable to or better than deep learning models such as ConvE. |
SciVQR: A Multidisciplinary Multimodal Benchmark for Advanced Scientific Reasoning Evaluation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing benchmarks for multimodal large language models fail to capture complexity and traceability of reasoning processes . SciVQR includes domain-specific visuals and challenges models to combine visual comprehension with reasoning. |
| Approach: | They propose a multimodal benchmark for scientific reasoning covering 54 subfields . SciVQR includes domain-specific visuals and challenges models to combine visual comprehension with reasoning . |
| Outcome: | SciVQR evaluates 54 subfields in mathematics, physics, chemistry, geography, astronomy, and biology . the results highlight the need for improved multi-step reasoning and integration of interdisciplinary knowledge . |
Compositional Structured Explanation Generation with Dynamic Modularized Reasoning (2024.starsem-1)
Copied to clipboard
| Challenge: | Large-scale language models have shown remarkable performance on reasoning tasks such as reading comprehension, natural language inference, story generation, etc. |
| Approach: | They propose a compositional structured explanation generation task to test a model's ability to generalize from generating entailment trees to more steps, focusing on the length and shapes of engorgement trees. |
| Outcome: | The proposed model shows competitive compositional generalization abilities in a generation setting. |
SIMPLOT: Enhancing Chart Question Answering by Distilling Essentials (2025.findings-naacl)
Copied to clipboard
| Challenge: | Recent advances in vision-language models have accelerated research into models capable of advanced reasoning based on images. |
| Approach: | They propose a method that leverages vision-language models to convert charts into table format . they use Large Language Model (LLM) for reasoning to extract only the essential information . |
| Outcome: | The proposed method extracts only the elements necessary for chart reasoning without the need for additional annotations or datasets. |
Seeing Beyond: Enhancing Visual Question Answering with Multi-Modal Retrieval (2025.coling-industry)
Copied to clipboard
| Challenge: | Multi-modal Large language models still suffer from model hallucination and lack of specific knowledge when answering challenging questions. |
| Approach: | They propose to use a multi-modal retrieval augmented generation method to integrate knowledge from all modalities into a model to enable alignment between query and knowledge. |
| Outcome: | The proposed method achieves significant performance improvement on the VQA dataset. |
RAGVUE: A Diagnostic View for Explainable and Automated Evaluation of Retrieval-Augmented Generation (2026.eacl-demo)
Copied to clipboard
| Challenge: | Existing tools for evaluating RAG systems often collapse heterogeneous behaviors into single scores. |
| Approach: | They propose a diagnostic framework for automated, reference-free evaluation of RAG pipelines. |
| Outcome: | The proposed framework decomposes RAG behavior into retrieval quality,answer relevance and completeness, strictclaim-level faithfulness, and judge calibration. |
CODI: Compressing Chain-of-Thought into Continuous Space via Self-Distillation (2025.emnlp-main)
Copied to clipboard
| 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. |
EscapeBench: Towards Advancing Creative Intelligence of Language Model Agents (2025.acl-long)
Copied to clipboard
Cheng Qian, Peixuan Han, Qinyu Luo, Bingxiang He, Xiusi Chen, Yuji Zhang, Hongyi Du, Jiarui Yao, Xiaocheng Yang, Denghui Zhang, Yunzhu Li, Heng Ji
| Challenge: | Existing language model agents excel in planning and reasoning, but lack creativity in unfamiliar environments. |
| Approach: | They propose a benchmark suite of room escape game environments to challenge agents with creative reasoning, unconventional tool use and iterative problem-solving to uncover implicit goals. |
| Outcome: | The proposed framework can perform with 40% fewer steps and hints and performs robustly across difficulty levels. |
BackMATH: Towards Backward Reasoning for Solving Math Problems Step by Step (2025.coling-industry)
Copied to clipboard
| Challenge: | Large language models (LLMs) have impressive results in reasoning, but when faced with more complex mathematical problems, performance drops significantly. |
| Approach: | They propose a backward reasoning dataset that includes 14K backward thinking problems and 100K reasoning steps. |
| Outcome: | The proposed model achieves an accuracy of 68.1% on the GSM8K dataset and 21.9% on the MATH dataset, exceeding the SOTA by 1.6% and 2.1% respectively. |
Test-Time Strategies for More Efficient and Accurate Agentic RAG (2026.acl-srw)
Copied to clipboard
Abhinav Sharma, Brian Zhang, Deepti Guntur, Zhiyang Zuo, Shreyas Chaudhari, Wenlong Zhao, Franck Dernoncourt, Puneet Mathur, Ryan A. Rossi, Nedim Lipka
| Challenge: | Retrieval-Augmented Generation (RAG) systems face challenges with complex, multi-hop questions. |
| Approach: | They propose to integrate contextualization module and de-duplication module to improve the accuracy of retrieved documents and to reduce the number of turns by 10.5%. |
| Outcome: | The proposed approach achieves a 5.6% increase in EM score and reduces the average number of turns by 10.5% compared to the baseline. |
First-Step Advantage: Importance of Starting Right in Multi-Step Math Reasoning (2025.findings-acl)
Copied to clipboard
| Challenge: | Language models can solve complex reasoning tasks better by learning to generate rationales for their predictions. |
| Approach: | They propose to use a larger model to guide smaller models to start . this allows them to generate rationales for their predictions when correct . |
| Outcome: | The proposed method improves performance on multistep reasoning datasets over multiple smaller models. |
Confidence Leaps in LLM Reasoning: Early Stopping and Cross-Model Transfer (2026.eacl-short)
Copied to clipboard
| Challenge: | Large Language Models build confidence gradually during reasoning, but internal dynamics of how confidence evolves during this reasoning process remain poorly understood. |
| Approach: | They propose a model-agnostic early-stopping heuristic that halts generation upon detecting a "confidence leap" they argue that conviction is often reached in a discrete "moment of insight" they propose to train models without sacrificing accuracy . |
| Outcome: | The proposed model-agnostic heuristic reduces generation time without sacrificing accuracy and significantly reduces the generation time. |
Nemotron-CrossThink: Scaling Self-Learning beyond Math Reasoning (2026.eacl-long)
Copied to clipboard
Syeda Nahida Akter, Shrimai Prabhumoye, Matvei Novikov, Seungju Han, Ying Lin, Evelina Bakhturina, Eric Nyberg, Yejin Choi, Mostofa Patwary, Mohammad Shoeybi, Bryan Catanzaro
| Challenge: | Prior work has successfully applied Reinforcement Learning (RL) to mathematical reasoning, but generalization to broader domains remains challenging due to limited data and lack of verifiable rewards for unstructured domains. |
| Approach: | They propose a framework that integrates multi-domain corpora into RL training to improve generalization across diverse reasoning tasks. |
| Outcome: | The proposed framework improves generalization across diverse reasoning tasks. |
A Layered Debating Multi-Agent System for Similar Disease Diagnosis (2025.naacl-short)
Copied to clipboard
| Challenge: | Traditional classification, contrastive learning, and large language models fail to detect subtle clues necessary for differentiation. |
| Approach: | They propose a framework that leverages Large Language Models to achieve accurate disease diagnosis . they structure patient information and integrate extensive medical knowledge to guide the analysis . |
| Outcome: | The proposed framework aims to identify subtle differences between similar diseases . the proposed framework can be used in clinical practice to improve accuracy . |
Can LLMs Learn From Mistakes? An Empirical Study on Reasoning Tasks (2024.findings-emnlp)
Copied to clipboard
| 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. |
Reasoning Like Program Executors (2022.emnlp-main)
Copied to clipboard
Xinyu Pi, Qian Liu, Bei Chen, Morteza Ziyadi, Zeqi Lin, Qiang Fu, Yan Gao, Jian-Guang Lou, Weizhu Chen
| Challenge: | Existing language models are inadequate in reasoning, according to studies . a new reasoning pre-training paradigm is based on pretraining language models with programs . |
| Approach: | They propose a reasoning pre-training paradigm that empowers language models to harvest reasoning knowledge possessed by program executors. |
| Outcome: | The proposed reasoning pre-training paradigm can boost models' reasoning skills . it can be instantiated by different kinds of program executors and run on a single database . |
TARo: Token-level Adaptive Routing for LLM Test-time Alignment (2026.findings-acl)
Copied to clipboard
Arushi Rai, Qiang Zhang, Hanqing Zeng, Yunkai Zhang, Dipesh Tamboli, Xiangjun Fan, Zhuokai Zhao, Lizhu Zhang
| Challenge: | Large language models (LLMs) exhibit strong reasoning capabilities but typically require expensive post-training to reach high performance. |
| Approach: | They propose to use token-level Adaptive Routing to steer frozen LLMs toward structured reasoning entirely at inference time. |
| Outcome: | Extensive experiments show that TARo significantly improves reasoning performance by up to +22.4% over base model and +8.4% . |
Explore the Reasoning Capability of LLMs in the Chess Testbed (2025.naacl-short)
Copied to clipboard
| Challenge: | a recent study shows that large language models struggle with long-term, complex reasoning tasks. |
| Approach: | They propose to integrate annotated strategy and tactic into large language models to improve reasoning capability. |
| Outcome: | The proposed model performs better than GPT, Claude, and Gemini models . it integrates annotated strategy and tactic into the model . |
Think Like You Execute: Verifiable Chain of Thought from Program Traces (2026.acl-industry)
Copied to clipboard
Shailja Thakur, Vaibhav Saxena, Rohan Kulkarni, Shivdeep Singh, Parameswaran Selvam, Hiroshi Kanayama, Hima Patel
| 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. |
Efficient Multi-Agent Collaboration with Tool Use for Online Planning in Complex Table Question Answering (2025.findings-naacl)
Copied to clipboard
| Challenge: | Recent studies on complex table question answering focus more and more on complex instances, as they are ubiquitous in table data analysis. |
| Approach: | They propose a framework that requires neither fine-tuning nor closed-source models to solve complex table question answering (TQA) their framework outperforms previous SoTA systems on three out of four benchmarks and performs comparably to the larger and more expensive closed-sourced model GPT-4 on two benchmarks. |
| Outcome: | The proposed framework outperforms closed-source models and closed-sourced models on three out of four benchmarks and performs comparable to the larger and more expensive closed-Source model GPT-4 on two benchmarks. |
How Well Do Multi-hop Reading Comprehension Models Understand Date Information? (2022.aacl-short)
Copied to clipboard
| Challenge: | Existing multi-hop reading comprehension datasets have reasoning shortcuts that can be used to answer comparison questions without performing multi- hop reasoning. |
| Approach: | They propose a dataset with three probing tasks in addition to the main question . they then evaluate the model's ability to understand date information . |
| Outcome: | The proposed model performs well in date comparison and number subtraction tasks. |
Evaluating the Performance of RAG Methods for Conversational AI in the Airport Domain (2025.naacl-industry)
Copied to clipboard
| Challenge: | a conversational AI system that answers standard airport queries and resolves airport terminology is ideal for airports from the top 20 in terms of annual passenger numbers. |
| Approach: | They propose a Conversational AI system that enables staff to communicate with flight systems . the system answers standard airport queries and resolves airport terminology . |
| Outcome: | The proposed system answers standard airport queries and resolves airport terminology, jargon, abbreviations and dynamic questions involving reasoning. |
Think Wider, Detect Sharper: Reinforced Reference Coverage for Document-Level Self-Contradiction Detection (2025.emnlp-main)
Copied to clipboard
| 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%. |
MapAgent: A Hierarchical Agent for Geospatial Reasoning with Dynamic Map Tool Integration (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing frameworks for large language models are tailored to domains such as mathematics, coding, or web automation. |
| Approach: | They propose a hierarchical multi-agent plug-and-play framework with customized toolsets and agentic scaffolds for map-integrated geospatial reasoning. |
| Outcome: | The proposed framework decouples planning from execution and reduces cognitive load on users. |
Chain-of-Thought Prompting Obscures Hallucination Cues in Large Language Models: An Empirical Evaluation (2025.findings-emnlp)
Copied to clipboard
| 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. |
S3Eval: A Synthetic, Scalable, Systematic Evaluation Suite for Large Language Model (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing benchmarks fail to evaluate extremely long-context LLMs or analyze their limitations. |
| Approach: | They propose a Synthetic, Scalable, Systematic evaluation suite for LLMs using SQL execution. |
| Outcome: | The proposed evaluation suite is able to scale text length and difficulty across scenarios and provides strong correlations with real-world benchmarks. |
Active Prompting with Chain-of-Thought for Large Language Models (2024.acl-long)
Copied to clipboard
| 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. |
DIAGRAMS : A Review Framework for Reasoning-Level Attribution in Diagram QA (2026.acl-demo)
Copied to clipboard
Anirudh Iyengar Kaniyar Narayana Iyengar, Tampu Ravi Kumar, Manan Suri, Raviteja Bommireddy, Dinesh Manocha, Puneet Mathur, Vivek Gupta
| Challenge: | Diagram question answering (Diagram QA) requires reasoning-level attribution that links each question-answer pair to all visual regions needed to derive the answer. |
| Approach: | They propose a diagram question-answer review framework that decouples interface logic from dataset-specific JSON structures through an internal meta-schema and dataset adapters. |
| Outcome: | The proposed framework achieves 85.39% precision and 75.30% recall across six diagram QA datasets. |
Efficient Reasoning for LLMs through Speculative Chain-of-Thought (2026.findings-acl)
Copied to clipboard
| 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. |
IndiFoodVQA: Advancing Visual Question Answering and Reasoning with a Knowledge-Infused Synthetic Data Generation Pipeline (2024.findings-eacl)
Copied to clipboard
| Challenge: | Large Vision Language Models lack domain-specific data for reasoning on complex problems. |
| Approach: | They propose to use explicit knowledge-infused questions, answers, and reasons to answer and reason upon the questions. |
| Outcome: | The proposed model improves by 25% over the baseline model. |
FlowVQA: Mapping Multimodal Logic in Visual Question Answering with Flowcharts (2024.findings-acl)
Copied to clipboard
Shubhankar Singh, Purvi Chaurasia, Yerram Varun, Pranshu Pandya, Vatsal Gupta, Vivek Gupta, Dan Roth
| Challenge: | Existing benchmarks for visual question answering lack in visual grounding and complexity, particularly in evaluating spatial reasoning skills. |
| Approach: | They propose to use flowcharts as visual contexts to assess the capabilities of visual question-answering multimodal language models in reasoning. |
| Outcome: | The proposed benchmarks evaluate models' ability to follow visual information without pre-existing knowledge on a suite of open-source and proprietary multimodal language models using various strategies, followed by an analysis of directional bias. |
DANLI: Deliberative Agent for Following Natural Language Instructions (2022.emnlp-main)
Copied to clipboard
Yichi Zhang, Jianing Yang, Jiayi Pan, Shane Storks, Nikhil Devraj, Ziqiao Ma, Keunwoo Yu, Yuwei Bao, Joyce Chai
| Challenge: | Recent work on embodied AI agents that can perform tasks by following human language instructions is limited by reactive methods, which are insufficient for long-horizon complex tasks. |
| Approach: | They propose a neuro-symbolic deliberative agent that, while following language instructions, proactively applies reasoning and planning based on its neural and symbolic representations acquired from past experience. |
| Outcome: | The proposed agent achieves greater than 70% improvement over reactive baselines on the challenging TEACh benchmark. |
FlowHN: Adaptive Token Routing for Efficient Parallel Hybrid Networks (2026.acl-industry)
Copied to clipboard
| Challenge: | Existing hybrids lack performance, latency, and cost-efficient scaling for production LLMs. |
| Approach: | They propose a deployment-oriented parallel hybrid architecture that enables deterministic conditional computation via FLOP-aware token circulation across attention and SSM branches. |
| Outcome: | FlowHN achieves 4 higher throughput and 15% higher MFU than current models while maintaining competitive accuracy on reasoning, coding, and long-context tasks. |
Disentangling Memory and Reasoning Ability in Large Language Models (2025.acl-long)
Copied to clipboard
Mingyu Jin, Weidi Luo, Sitao Cheng, Xinyi Wang, Wenyue Hua, Ruixiang Tang, William Yang Wang, Yongfeng Zhang
| Challenge: | Existing LLMs operate as an opaque process without explicit separation between knowledge retrieval and reasoning steps, making the decision-making process unclear and disorganized. |
| Approach: | They propose a language model inference paradigm that decomposes the complex inference process into two distinct and clear actions: (1) memory recall: which retrieves relevant knowledge, and (2) reasoning: which performs reasoning steps based on the recalled knowledge. |
| Outcome: | The proposed paradigm decomposes the inference process into two distinct and clear actions, memory and reason, guiding the model to distinguish between steps that require knowledge retrieval and those that involve reasoning. |
Modeling Temporal-Modal Entity Graph for Procedural Multimodal Machine Comprehension (2022.acl-long)
Copied to clipboard
| Challenge: | Procedural Multimodal Documents organize textual instructions and corresponding images step by step. |
| Approach: | They propose a novel temporal-modal entity Graph for comprehending PMDs . they propose encoding and reasoning modules to capture textual and visual entities . |
| Outcome: | The proposed model can capture textual and visual entities and trace their temporal-modal evolution. |
See More, Store Less: Memory-Efficient Resolution for Video Moment Retrieval (2026.findings-eacl)
Copied to clipboard
| Challenge: | Existing video moment retrieval methods rely on sparse frame sampling, risking information loss. |
| Approach: | a new video-based framework enhances memory efficiency while maintaining high information resolution . SMORE uses query-guided captions to encode semantics aligned with user intent . |
| Outcome: | a new framework improves memory efficiency while maintaining high information resolution . it achieves state-of-the-art performance on QVHighlights, Charades-STA, and ActivityNet-Captions benchmarks . |
Ada-RS: Adaptive Rejection Sampling for Selective Thinking (2026.acl-industry)
Copied to clipboard
Yirou Ge, Yixi Li, Alec M. Chiu, Shivani Shekhar, Zijie Pan, Avinash Thangali, Yun-Shiuan Chuang, Chaitanya Kulkarni, Uma Kona, Linsey Pang, Prakhar Mehrotra
| Challenge: | Large language models are increasingly being deployed in cost- and latency-sensitive settings . chain-of-thought improves reasoning, but it can waste tokens on simple requests . |
| Approach: | They introduce an algorithm-agnostic sample filtering framework for learning selective reasoning . they show that Ada-RS reduces average output tokens by 80% and reducing thinking rate by 5% . |
| Outcome: | The proposed framework reduces output tokens by 80% and thinking rate by 95% on a synthetic tool call-oriented e-commerce benchmark. |
Garbage In, Reasoning Out? Why Benchmark Scores are Unreliable and What to Do About It (2026.findings-eacl)
Copied to clipboard
| Challenge: | Using social reasoning benchmarks, we uncover pervasive flaws in both benchmark items and evaluation methodology. |
| Approach: | They audit three widely used social reasoning benchmarks and identify flaws in their design and evaluation methodology. |
| Outcome: | The results challenge the validity of current benchmark-based claims about social reasoning in large language models. |
Critic-CoT: Boosting the Reasoning Abilities of Large Language Model via Chain-of-Thought Critic (2025.findings-acl)
Copied to clipboard
Xin Zheng, Jie Lou, Boxi Cao, Xueru Wen, Yuqiu Ji, Hongyu Lin, Yaojie Lu, Xianpei Han, Debing Zhang, Le Sun
| Challenge: | Existing approaches to improve the reasoning performance of large language models rely on intuitive instance-level feedback, which limits the reasoning capabilities. |
| Approach: | They propose a framework that pushes LLMs toward System-2-like critic capability by using a step-wise CoT reasoning paradigm and automatic construction of weak-supervision data without human annotation. |
| Outcome: | The proposed model significantly improves task-solving performance by filtering out invalid solutions or iterative refinement. |
Thinking with DistilQwen: A Tale of Four Distilled Reasoning and Reward Model Series (2025.emnlp-industry)
Copied to clipboard
| Challenge: | In the rapidly evolving landscape of large language models, the need for efficient reasoning models has become increasingly urgent. |
| Approach: | They extend the Qwen model family by introducing four model series specifically designed for industrial applications. |
| Outcome: | The proposed models outperform previous models in multiple benchmarks and provide scalable training and inference functionality on the Alibaba Cloud PAI platform. |
Beyond Token Length: Step Pruner for Efficient and Accurate Reasoning in Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing reinforcement learning methods for large reasoning models suffer from excessive verbosity, known as "overthinking." Existing models penalize generated tokens to promote conciseness, but these methods encounter two challenges: they may develop hacking behavior in later stages of training by discarding reasoning steps. |
| Approach: | They propose a framework that steers large reasoning models toward more efficient reasoning . they prioritize correctness while imposing penalties for redundant steps . |
| Outcome: | The proposed framework reduces token usage by 69.7% on AIME24. |
MenatQA: A New Dataset for Testing the Temporal Comprehension and Reasoning Abilities of Large Language Models (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models (LLMs) have shown nearly saturated performance on many NLP tasks. |
| Approach: | They construct multiple sensitive factors time QA which encompasses three temporal factors . they test current mainstream LLMs with different parameter sizes . |
| Outcome: | The proposed model incorporates three temporal factors with 2,853 samples . the results show that LLMs fall behind smaller models on these factors . |
MARCO: Multi-Agent Real-time Chat Orchestration (2024.emnlp-industry)
Copied to clipboard
Anubhav Shrimal, Stanley Kanagaraj, Kriti Biswas, Swarnalatha Raghuraman, Anish Nediyanchath, Yi Zhang, Promod Yenigalla
| Challenge: | MARCO is a multi-agent real-time chat orchestration framework for automating workflows that require interactions with tools, reasoning, and human collaboration. |
| Approach: | They propose a multi-agent real-time chat orchestration framework for automating workflows using LLMs. |
| Outcome: | The proposed framework performs with 94.48% accuracy and 92.74% accuracy on restaurant and retail conversations datasets and 44.91% improved latency and 33.71% cost reduction in a production setting. |
GEMMAS: Graph-based Evaluation Metrics for Multi Agent Systems (2025.emnlp-industry)
Copied to clipboard
| Challenge: | Existing evaluations focus on the correctness of the final output, overlooking inefficient communication and poor coordination contribute to redundant reasoning and higher computational costs. |
| Approach: | They propose a graph-based evaluation framework that analyzes the internal collaboration process by modeling agent interactions as a directed acyclic graph. |
| Outcome: | The proposed framework shows that outcome-only metrics are insufficient for evaluating multi-agent performance on GSM8K. |
Reasoning or Knowledge: Stratified Evaluation of Biomedical LLMs (2026.eacl-long)
Copied to clipboard
| Challenge: | Medical reasoning in large language models is a complex cognitive process through which clinicians interpret patient data and make diagnostic and therapeutic decisions. |
| Approach: | They propose an evaluation framework that disentangles knowledge recall from reasoning by training a PubMedBERT-based classifier and applying it to 11 widely used biomedical QA benchmarks. |
| Outcome: | The proposed evaluation framework disentangles knowledge recall from reasoning by training a PubMedBERT-based classifier and applying it to 11 widely used biomedical QA benchmarks. |
Enhancing Advanced Visual Reasoning Ability of Large Language Models (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in Vision-Language (VL) research have sparked new benchmarks for complex visual reasoning, challenging models’ advanced reasoning ability. |
| Approach: | They propose a novel multi-modal in-context learning methodology to enhance LLMs’ contextual understanding and reasoning. |
| Outcome: | The proposed model achieves SOTA performance among all visual reasoning tasks and achieves a 'higher level of accuracy' than previous models. |
Dynamic Relevance Graph Network for Knowledge-Aware Question Answering (2022.coling-1)
Copied to clipboard
| Challenge: | Existing approaches to solve commonsense question answering problems often miss some edges between entities, which breaks the reasoning chain. |
| Approach: | They propose a graph neural network architecture that uses relevance as graph edges to establish new edges dynamically for learning node representations in the graph network. |
| Outcome: | The proposed approach shows competitive performance on two QA benchmarks, CommonsenseQA and OpenbookQA, compared to the state-of-the-art published results. |
AutoMonitor-Bench: Evaluating the Reliability of LLM-Based Misbehavior Monitor (2026.findings-acl)
Copied to clipboard
| Challenge: | AutoMonitor-Bench evaluates the reliability of LLM-based misbehavior monitors across diverse tasks and failure modes. |
| Approach: | They introduce AutoMonitor-Bench, a benchmark designed to evaluate misbehavior monitors across diverse tasks and failure modes. |
| Outcome: | The new benchmark evaluates the reliability of LLM-based misbehavior monitors across tasks and failure modes. |
AWARE: Agentic Knowledge Warehousing for Contextual Intelligence (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models excel in information seeking tasks, but their knowledge is limited in coverage and timeliness. |
| Approach: | They propose an agentic knowledge warehousing framework that transforms unstructured data into minimal, task-conditioned knowledge representations consumable by LLMs. |
| Outcome: | Experiments on GAIA, WebWalker, and BrowseComp-Plus show improvements over baselines. |
Exploiting Contextual Information via Dynamic Memory Network for Event Detection (D18-1)
Copied to clipboard
| Challenge: | Existing methods for event detection only process context once . a multi-hop mechanism to capture contextual information improves performance . |
| Approach: | They propose to use dynamic memory network to capture contextual information . they propose to model event triggers by identifying word or phrase which most represents it . |
| Outcome: | The proposed model achieves best F1 score compared to the state-of-the-art models. |
MathMist: A Parallel Multilingual Benchmark Dataset for Mathematical Problem Solving and Reasoning (2026.findings-eacl)
Copied to clipboard
Mahbub E Sobhani, Md. Faiyaz Abdullah Sayeedi, Tasnim Mohiuddin, Md Mofijul Islam, Swakkhar Shatabda
| 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. |
Toward Inclusive Language Models: Sparsity-Driven Calibration for Systematic and Interpretable Mitigation of Social Biases in LLMs (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a new method to mitigate stereotypical bias in large language models is needed . inherent biases from training on vast Internet datasets can amplify harmful stereotypes . |
| Approach: | They propose a method to identify stereotypical bias in decoder-only transformer models . they apply a localization mechanism that correlates internal activations with a new Context Influence score . |
| Outcome: | The proposed method reduces stereotypical biases on BBQ, StereoSet, and CrowS-Pairs while improving reasoning performance on MMLU by 10%. |
DentalGPT: Incentivizing Multimodal Reasoning in Dentistry (2026.findings-acl)
Copied to clipboard
Zhenyang Cai, Jiaming Zhang, Junjie Zhao, Ziyi Zeng, Yanchao Li, Liang Jingyi, Junying Chen, Yunjin Yang, Jiajun You, Shuzhi Deng, null Xieruiqiii, Yuanting Chen, Xiangyi Feng, Jianquan Li, Liangyi Chen, Junwen Wang, Shan Jiang, Benyou Wang
| Challenge: | Current multimodal large language models (MLLMs) show limited understanding of dental images. |
| Approach: | They propose a dental-specialized multimodal large language model trained via staged multimodal alignment and reinforcement learning. |
| Outcome: | The proposed model outperforms state-of-the-art models on disease classification and dental VQA tasks. |
Two Heads Are Better Than One: Dual-Model Verbal Reflection at Inference-Time (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing LLMs struggle to reliably detect subtle reasoning errors in ASAS tasks. |
| Approach: | They propose a dual-model framework with a dedicated Critic model trained for effective reflection that generates precise verbal feedback. |
| Outcome: | The proposed framework outperforms existing ASAS benchmarks and provides valuable insights into the performance of the proposed framework. |
LLMs Are Not Intelligent Thinkers: Introducing Mathematical Topic Tree Benchmark for Comprehensive Evaluation of LLMs (2025.naacl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) have impressive capabilities in mathematical reasoning, but their effectiveness is limited to specific mathematical topics. |
| Approach: | They propose to use the MaTT benchmark to assess large language models' accuracy in multiple-choice scenarios. |
| Outcome: | The proposed model achieved 54% accuracy in a multiple-choice scenario, while the Chain-of-Thought prompting did not improve. |
Optimizing Deeper Transformers on Small Datasets (2021.acl-long)
Copied to clipboard
Peng Xu, Dhruv Kumar, Wei Yang, Wenjie Zi, Keyi Tang, Chenyang Huang, Jackie Chi Kit Cheung, Simon J.D. Prince, Yanshuai Cao
| Challenge: | a common belief that training deep transformers from scratch requires large datasets is wrong . however, with proper initialization and optimization, the benefits of very deep transformer can carry over to challenging tasks with small datasets. |
| Approach: | They train 48 layers of transformers from pre-trained RoBERTa and 24 relation-aware layers from scratch. |
| Outcome: | The proposed scheme achieves state-of-the-art performance on a text-to-sql parsing benchmark . it uses 24 fine-tuned layers from pre-trained RoBERTa and 24 relation-aware layers from scratch . |
Is This LLM Library Learning? Evaluation Must Account For Compute and Behaviour (2026.eacl-long)
Copied to clipboard
| Challenge: | Recent advances in the coding, reasoning, and tool-use ability of LLMs have raised the possibility of library learning with LLM. |
| Approach: | They propose to use reusable and composable functions and tools to create reusable, composesable code and tools that can be reused by modifying relevant examples. |
| Outcome: | The proposed system fails to consistently outperform the baseline model and does not correct for the difference in computational cost. |
Tell-and-Answer: Towards Explainable Visual Question Answering using Attributes and Captions (D18-1)
Copied to clipboard
| Challenge: | Existing approaches to visual question answering represent images using pre-trained CNNs . but they rarely provide any insight, apart from the answer, into the VQA process . |
| Approach: | They propose to break up the end-to-end VQA into two steps: explaining and reasoning . they first extract attributes and generate descriptions as explanations for an image . a reasoning module utilizes these explanations in place of the image to infer an answer . |
| Outcome: | The proposed system achieves comparable performance with baselines, but with added benefits of explanability and the ability to improve with higher quality explanations. |
Direct Evaluation of Chain-of-Thought in Multi-hop Reasoning with Knowledge Graphs (2024.findings-acl)
Copied to clipboard
| 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. |
Don’t Tell the Answer, Truly Guide the Reasoning During RL Rollouts (2026.findings-acl)
Copied to clipboard
Xinyi Wang, Jinyi Han, Zishang Jiang, Tingyun li, Jiaqing Liang, Sihang Jiang, Zhaoqian Dai, Ma Shuguang, Fei Yu, Yanghua Xiao
| 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. |
A Compliance Checking Framework Based on Retrieval Augmented Generation (2025.coling-main)
Copied to clipboard
| Challenge: | Existing text-based compliance checking methods are limited by their flexibility and lack structure. |
| Approach: | They propose a text-based compliance checking framework based on Retrieval-Augmented Generation that integrates a static layer for storing factual knowledge, a dynamic layer for retrieval and reasoning, and an eventic graph to structurally describe regulatory information. |
| Outcome: | The proposed framework consistently achieves state-of-the-art results across various scenarios surpassing baselines. |
ReaRev: Adaptive Reasoning for Question Answering over Knowledge Graphs (2022.findings-emnlp)
Copied to clipboard
| Challenge: | Knowledge Graph Question Answering (KGQA) involves retrieving entities as answers from a Knowledge Flow using natural language queries. |
| Approach: | They propose a method to decode a question into instructions that are dense question representations used to guide the KG traversals. |
| Outcome: | The proposed method improves instruction decoding and execution by using a KG-aware information to update the initial instructions. |
CLINE: Contrastive Learning with Semantic Negative Examples for Natural Language Understanding (2021.acl-long)
Copied to clipboard
| Challenge: | Pre-trained language models are vulnerable to simple perturbations, causing poor robustness . recent studies show that adversarial training is useless or harmful for the model to detect these semantic changes. |
| Approach: | They propose to use adversarial training to improve the robustness of pre-trained models . they propose to construct negative examples with similar and opposite semantics . |
| Outcome: | Empirical results show that the proposed approach improves on sentiment analysis, reasoning, and reading comprehension tasks. |
Document-level Biomedical Relation Extraction Based on Multi-Dimensional Fusion Information and Multi-Granularity Logical Reasoning (2022.coling-1)
Copied to clipboard
| Challenge: | Existing models with reasoning are single-granularity based on one element information, ignoring complementary fact of different granularities. |
| Approach: | They propose a document-level biomedical relation extraction model called FILR . it uses multi-dimensional information fusion and multi-granularity logic to obtain rich inferences . |
| Outcome: | The proposed model extracts all relation facts from biomedical documents . it is based on multi-dimensional information fusion and multi-granularity logic reasoning . the proposed model achieves state-of-the-art performance on two widely used biomedically corpora . |
ClarET: Pre-training a Correlation-Aware Context-To-Event Transformer for Event-Centric Generation and Classification (2022.acl-long)
Copied to clipboard
| Challenge: | Existing work on event-centric reasoning fails to model event-level correlations . Existing studies limit their scope to specific scenarios or overlook event- level correlations. |
| Approach: | They propose to pre-train a general Correlation-aware context-to-Event Transformer for event-centric reasoning by highlighting event-level correlations with effective training. |
| Outcome: | The proposed model is applicable to a wide range of event-centric reasoning scenarios, considering its versatility of event correlation types, application formulations, and reasoning types. |
DocAssistant: Integrating Key-region Reading and Step-wise Reasoning for Robust Document Visual Question Answering (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing document understanding models focus on key information and generate answers straightforwardly . existing models ignore evidence from source documents and lack interpretability . |
| Approach: | They propose a visual encoder that fuses text into visual encoded visual encodes . they use multimodal large language models as data generators and checkers to generate step-wise question-and-answer pairs for document images. |
| Outcome: | The proposed model can answer step-wise questions without compromising the performance of the original model. |
Learning SQL Like a Human: Structure-Aware Curriculum Learning for Text-to-SQL Generation (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing models struggle with complex queries, especially multi-table joins and reasoning. |
| Approach: | They propose to build a model with synthetic training samples and a structure-aware curriculum learning framework for enhancing SQL generation. |
| Outcome: | The proposed model improves on the existing model on the Spider and Bird benchmarks. |
Reasoning’s Razor: Reasoning Improves Accuracy but Hurts Recall at Critical Operating Points in Safety and Hallucination Detection (2026.eacl-long)
Copied to clipboard
Atoosa Chegini, Hamid Kazemi, Garrett Souza, Maria Safi, Yang Song, Samy Bengio, Sinead Williamson, Mehrdad Farajtabar
| Challenge: | a new study examines the suitability of reasoning for precision-sensitive classification tasks . false positives carry severe operational consequences, such as blocking legitimate queries . |
| Approach: | They propose to use reasoning for classification tasks under low false positive rate regimes . they find that Think On improves overall accuracy, but performs poorly at low FPRs a . |
| Outcome: | The proposed reasoning-augmented generation model outperforms self-verbalized confidence in precision-sensitive deployments. |
Unveiling Confirmation Bias in Chain-of-Thought Reasoning (2025.findings-acl)
Copied to clipboard
| 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. |
DRP: Distilled Reasoning Pruning with Mathematical Skill-aware Step Decomposition for Efficient Large Reasoning Models (2026.findings-acl)
Copied to clipboard
| 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. |
Meta-Reflection: A Feedback-Free Reflection Learning Framework (2025.acl-long)
Copied to clipboard
Yaoke Wang, Yun Zhu, XintongBao XintongBao, Wenqiao Zhang, Suyang Dai, Kehan Chen, Wenqiang Li, Gang Huang, Siliang Tang, Yueting Zhuang
| Challenge: | Existing approaches to improve large language models' ability to understand and reason are limited by external feedback. |
| Approach: | They propose a feedback-free reflection mechanism that requires only a single inference pass without external feedback. |
| Outcome: | The proposed method is based on an industrial e-commerce benchmark and public datasets. |
Reasoning with Trees: Faithful Question Answering over Knowledge Graph (2025.coling-main)
Copied to clipboard
| Challenge: | Recent advances in large language models (LLMs) have shown remarkable progress in reasoning capabilities, yet they still face challenges in complex, multi-step reasoning tasks. |
| Approach: | They propose a framework that synergistically integrates LLMs with knowledge graphs (KGs) to enhance reasoning performance and interpretability. |
| Outcome: | The proposed framework outperforms existing state-of-the-art methods on two benchmark KGQA datasets and improves on the MCTS process. |
Deciphering the Factors Influencing the Efficacy of Chain-of-Thought: Probability, Memorization, and Noisy Reasoning (2024.findings-emnlp)
Copied to clipboard
| 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. |
Self-contradictory reasoning evaluation and detection (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models (LLMs) have shown impressive reasoning ability, but many downstream reasoning tasks focus on performance-wise evaluation. |
| Approach: | They define and assess the Self-Contra rate across three datasets and delve into finer-grained categories of Self-contra reasoning. |
| Outcome: | The proposed model can detect self-contra reasoning with a 52.2% F1 score, much lower than for humans. |
D2Plan: Dual-Agent Dynamic Global Planning for Complex Retrieval-Augmented Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Recent advances in reinforcement learning (RL) have empowered Large Language Models (LLMs) with the capability to perform autonomous retrieval during reasoning tasks. |
| Approach: | They propose a "D2Plan" paradigm for retrieval-augmented reasoning that integrates a 'Reasoner' and a'Purifier' |
| Outcome: | Experiments show that the proposed paradigm improves on QA benchmarks. |
RepoDistill: Distilling Repository Knowledge through Compression-Aware Budget Allocation and Policy Optimization (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have strong performance on code translation tasks, but they struggle with repository-level scenarios where context is extensive and interdependent. |
| Approach: | They propose a framework that integrates retrieval with learning budget allocation for fine-grained context compression. |
| Outcome: | The proposed framework outperforms baselines on SWE-QA, CoderEval, and LongCodeU. |
Ruleformer: Context-aware Rule Mining over Knowledge Graph (2022.coling-1)
Copied to clipboard
| Challenge: | Existing work on rule mining focuses on mining rules, but how to select appropriate rules for completion of different triplets has not been discussed. |
| Approach: | They propose to take context information into consideration when selecting suitable rules . they devise a transformer-based rule mining approach, Ruleformer . |
| Outcome: | The proposed model takes context information into consideration, which helps select suitable rules for inference tasks. |
Right for the Right Reason: Evidence Extraction for Trustworthy Tabular Reasoning (2022.acl-long)
Copied to clipboard
| Challenge: | Recent studies show that tabular reasoning models use spurious correlations and focus on false evidence or ignore it altogether. |
| Approach: | They propose a task where models need to extract evidence and then inference labels . they crowdsource evidence row labels and develop unsupervised evidence extraction strategies . |
| Outcome: | The proposed approach outperforms baseline models on the inference task using only the automatically extracted evidence as the premise. |
Curriculum: A Broad-Coverage Benchmark for Linguistic Phenomena in Natural Language Understanding (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing evaluation methods do not provide insight into how well a language model captures distinct linguistic skills essential for language understanding and reasoning. |
| Approach: | They propose a new format of NLI benchmark for evaluation of broad-coverage linguistic phenomena using a set of datasets and an evaluation procedure for diagnosing how well a language model captures reasoning skills. |
| Outcome: | The proposed model can diagnose model behavior and verify model learning quality. |
When Reasoning Meets Information Aggregation: A Case Study with Sports Narratives (2024.emnlp-main)
Copied to clipboard
| Challenge: | Using sports data, an LLM can analyze sports narratives to infer points from actions, identify related entities, attribute points accurately to players and teams, and draw conclusions. |
| Approach: | They propose a method to synthesize NBA basketball game narratives using real NBA basketball data and propose 'SportsGen' they find that most models fail to accurately aggregate basketball scores due to frequent scoring patterns and open-source models suffer from significant score hallucinations. |
| Outcome: | The proposed method can evaluate LLMs’ reasoning capabilities under complex scenarios with varying narrative lengths and density of information. |
How Well Do LLMs Handle Cantonese? Benchmarking Cantonese Capabilities of Large Language Models (2025.findings-naacl)
Copied to clipboard
| Challenge: | Cantonese has scant representation in NLP research, especially compared to other languages from similarly developed regions. |
| Approach: | They propose to evaluate Cantonese LLM performance in factual generation, mathematical logic, complex reasoning, and general knowledge in Cantonesian. |
| Outcome: | The proposed models will evaluate Cantonese's performance in factual generation, mathematical logic, complex reasoning, and general knowledge in Cantone. |
HotpotQA: A Dataset for Diverse, Explainable Multi-hop Question Answering (D18-1)
Copied to clipboard
Zhilin Yang, Peng Qi, Saizheng Zhang, Yoshua Bengio, William Cohen, Ruslan Salakhutdinov, Christopher D. Manning
| Challenge: | Existing question answering (QA) datasets fail to train QA systems to perform complex reasoning and provide explanations for answers. |
| Approach: | They propose a new dataset with 113k Wikipedia-based question-answer pairs with four key features: (1) the questions require finding and reasoning over multiple supporting documents to answer; (2) the questions are diverse and not constrained to any pre-existing knowledge bases or knowledge schemas; (3) the questions provide sentence-level supporting facts required for reasoning; and (4) a type of factoid comparison questions to test QA systems’ ability to extract relevant facts and perform necessary comparison. |
| Outcome: | The proposed dataset has 113k Wikipedia-based question-answer pairs and four key features that make it challenging for the latest QA systems. |
Representation Learning with Ordered Relation Paths for Knowledge Graph Completion (D19-1)
Copied to clipboard
| Challenge: | Existing knowledge graphs are incomplete and lack the order of relations in paths. |
| Approach: | They propose a method which takes relation paths into account but ignores order of relations in paths which is important for reasoning. |
| Outcome: | The proposed method performs better than state-of-the-art methods on two benchmark datasets. |
Asynchronous Deep Interaction Network for Natural Language Inference (D19-1)
Copied to clipboard
| Challenge: | Existing methods have framed the reasoning problem as a semantic matching task. |
| Approach: | They propose an asynchronous deep interaction network (ADIN) to deconstruct the reasoning process and implement asynchron and multi-step reasoning. |
| Outcome: | The proposed model outperforms strong baselines on three popular benchmarks: SNLI, MultiNLI, and SciTail. |
Hello Again! LLM-powered Personalized Agent for Long-term Dialogue (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing dialogue systems focus on brief single-session interactions, neglecting real-world needs for long-term companionship and personalized interactions. |
| Approach: | They propose a model-agnostic framework for long-term dialogue agents . they use event summary and persona management to enable reasoning . |
| Outcome: | The proposed framework incorporates three independently tunable modules dedicated to event perception, persona extraction, and response generation. |
Tracking the Limits of Knowledge Propagation: How LLMs Fail at Multi-Step Reasoning with Conflicting Knowledge (2026.eacl-long)
Copied to clipboard
| Challenge: | Existing benchmarks for analyzing the performance of Large Language Models (LLMs) focus on single knowledge updates and fact recall, but do not consider how these updates affect downstream reasoning. |
| Approach: | They propose a benchmark to study how LLMs propagate new knowledge when it conflicts with the model's parametric knowledge. |
| Outcome: | The proposed benchmark compared models with no updated facts to show that the new methods worsen performance and improve reasoning performance. |
ClaimPKG: Enhancing Claim Verification via Pseudo-Subgraph Generation with Lightweight Specialized LLM (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing verification methods rely on unstructured text corpora to break down claims . despite strong reasoning abilities, modern LLMs struggle with modular pipelines . |
| Approach: | They propose a framework that integrates knowledge graphs with LLM reasoning . they propose KGs provide structured, semantically rich representations . |
| Outcome: | The proposed framework outperforms baselines on the FactKG dataset by 9%-12% accuracy points across multiple categories. |
Semantic Parsing for Technical Support Questions (C18-1)
Copied to clipboard
| Challenge: | Automating technical support is a task of providing answers to complex problems . traditional approaches to this task rely on information retrieval and are keyword based . |
| Approach: | They propose a semantic parsing approach that uses grammatical structure to extract technical questions' attributes as a baseline and a CRF-based model that can improve performance in the presence of annotated data. |
| Outcome: | The proposed model outperforms retrieval baselines in annotated data for training. |
Humans Keep It One Hundred: an Overview of AI Journey (2020.lrec-1)
Copied to clipboard
Tatiana Shavrina, Anton Emelyanov, Alena Fenogenova, Vadim Fomin, Vladislav Mikhailov, Andrey Evlampiev, Valentin Malykh, Vladimir Larin, Alex Natekin, Aleksandr Vatulin, Peter Romov, Daniil Anastasiev, Nikolai Zinov, Andrey Chertok
| Challenge: | Artificial General Intelligence (AGI) is showing growing performance in numerous applications - beating human performance in Chess and Go, using knowledge bases and text sources to answer questions and even pass human examination. |
| Approach: | They propose to use knowledge bases and text sources to answer questions to improve AI performance on knowledge bases, reasoning and text generation. |
| Outcome: | The proposed AI Journey system passed the final native language exam in Russian with a high score of 69%, with 68% being an average human result. |
Structured Episodic Event Memory (2026.acl-long)
Copied to clipboard
| Challenge: | Current approaches to memory in Large Language Models (LLMs) rely on static Retrieval-Augmented Generation (RAG) this lacks the cognitive organization necessary to model the dynamic and associative nature of long-term interaction. |
| Approach: | They propose a hierarchical framework that transforms interaction streams into structured Episodic Event Frames (EEFs) anchored by precise provenance pointers. |
| Outcome: | The proposed framework outperforms baseline approaches on LoCoMo and LongMemEval benchmarks. |
Enhancing Zero-shot and Few-shot Stance Detection with Commonsense Knowledge Graph (2021.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for stance detection are not applicable to zero-shot and few-shot scenarios. |
| Approach: | They propose a model that integrates commonsense knowledge into a stance detection model. |
| Outcome: | The proposed model outperforms state-of-the-art methods on zero-shot and few-shot stance detection tasks. |
Improve Rule Retrieval and Reasoning with Self-Induction and Relevance ReEstimate (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing rule retrieval methods suffer from low accuracy due to semantic gap between instantiated facts and abstract representations of rules. |
| Approach: | They propose a method that induces inferential rules that might offer benefits for reasoning by abstracting the underlying knowledge and logical structure in queries. |
| Outcome: | The proposed method improves retrieval effectiveness and accuracy across settings. |
Hierarchical Deconstruction of LLM Reasoning: A Graph-Based Framework for Analyzing Knowledge Utilization (2024.emnlp-main)
Copied to clipboard
| Challenge: | Despite advances in large language models, how they use their knowledge for reasoning is not yet well understood. |
| Approach: | They propose a method that deconstructs complex real-world questions into a graph . they quantify forward discrepancy, a discrepany in LLM performance on simpler sub-problems . |
| Outcome: | The proposed method shows that smaller models exhibit more discrepancies than larger models . it also shows that guiding models from simpler to complex questions improves performance . |
VCD: A Dataset for Visual Commonsense Discovery in Images (2025.findings-acl)
Copied to clipboard
| Challenge: | Visual commonsense data sets lack visual grounded representations of commonsensense . existing knowledge bases lack visual-based knowledge tied to actual visual scenes . |
| Approach: | They present a large-scale visual commonsense dataset with over 100,000 images and 14 million object-commonsense pairs that integrates both Seen (directly observable) and Unseen (inferrable) commonsens. |
| Outcome: | The proposed model integrates Seen (directly observable) and Unseen (inferrable) commonsense across Property, Action, and Space aspects. |
Reasoning about Ambiguous Definite Descriptions (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing resources to evaluate reasoning are not well suited to investigate the capability of resolving ambiguities by explicit reasoning. |
| Approach: | They propose to use ambiguous definite descriptions to create a benchmark dataset which requires models to resolve ambiguity by explicit reasoning. |
| Outcome: | The proposed model includes all information required to resolve the ambiguity in the prompt, which means a model does not require anything but reasoning to do well. |
Is a Question Decomposition Unit All We Need? (2022.emnlp-main)
Copied to clipboard
| Challenge: | Large Language Models (LMs) have achieved state-of-the-art performance on many NLP benchmarks. |
| Approach: | They propose to decompose a hard question into simpler questions that are easier for models to answer. |
| Outcome: | The proposed approach significantly improves model performance (24% for GPT3 and 29% for RoBERTa-SQuAD along with a symbolic calculator) by decomposing a hard question into simpler questions that are easier for models to answer. |
LATENTLOGIC: Learning Logic Rules in Latent Space over Knowledge Graphs (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for learning logic rules for knowledge graph reasoning face limitations such as searching in vast search space and inefficient optimization. |
| Approach: | They propose a framework to efficiently mine logic rules by controllable generation in the latent space by a pre-trained VAE and a discriminator. |
| Outcome: | The proposed framework efficiently mines logic rules by controllable generation in the latent space. |
Disentangling Biased Knowledge from Reasoning in Large Language Models via Machine Unlearning (2025.acl-long)
Copied to clipboard
Zheyuan Liu, Suraj Maharjan, Fanyou Wu, Rahil Parikh, Belhassen Bayar, Srinivasan H. Sengamedu, Meng Jiang
| Challenge: | Existing approaches to disentangle biased knowledge from reasoning are sub-optimal . entangled data makes curation difficult, leading to inclusion of sensitive, toxic data. |
| Approach: | They propose a framework that selectively removes biased knowledge while preserving reasoning abilities. |
| Outcome: | The proposed framework improves fairness accuracy by 14.7% and reasoning performance by 62.6% across multiple LLMs. |
Marginal Likelihood Training of BiLSTM-CRF for Biomedical Named Entity Recognition from Disjoint Label Sets (D18-1)
Copied to clipboard
| Challenge: | Existing large labeled text datasets contain labels for multiple subsets of biomedical entity types, but it is rare to find large labeling datasets containing all desired entity types together. |
| Approach: | They propose a method for training a single CRF extractor from multiple datasets with disjoint or partially overlapping sets of entity types. |
| Outcome: | The proposed method improves NER F1 over training in isolation on biocreative V CDR, biocreativ VI ChemProt and MedMentions datasets. |
Rethinking Offensive Text Detection as a Multi-Hop Reasoning Problem (2022.findings-acl)
Copied to clipboard
| Challenge: | Existing methods of offensive text detection perform poorly when asked to detect implicitly offensive statements . a dataset based on SLIGHT provides a framework for implicit offensive text identification . |
| Approach: | They propose a dataset to support the task of implicit offensive text detection in dialogues . they show that reasoning is crucial for understanding this broader class of offensive utterances - SLIGHT . |
| Outcome: | The proposed model achieves 11% accuracy in implicit offensive text detection tasks . the proposed model can be used to identify toxic speech in specific domains . |
CipherBank: Exploring the Boundary of LLM Reasoning Capabilities through Cryptography Challenge (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) have demonstrated remarkable capabilities, but their capabilities in cryptographic decryption tasks remain underexplored. |
| Approach: | They propose a benchmark to evaluate the reasoning capabilities of large language models in cryptographic decryption tasks. |
| Outcome: | The proposed benchmark examines the reasoning capabilities of large language models in cryptographic decryption tasks. |
Retrieving Multimodal Information for Augmented Generation: A Survey (2023.findings-emnlp)
Copied to clipboard
Ruochen Zhao, Hailin Chen, Weishi Wang, Fangkai Jiao, Xuan Long Do, Chengwei Qin, Bosheng Ding, Xiaobao Guo, Minzhi Li, Xingxuan Li, Shafiq Joty
| Challenge: | Large Language Models (LLMs) are increasingly using multimodality to augment their generation ability, but there is no unified perception of at which stage and how to incorporate different modalities. |
| Approach: | They propose to use multimodality to augment Large Language Models (LLMs) this will provide scholars with a deeper understanding of the methods' applications and encourage them to adapt existing techniques to the fast-growing field of LLMs. |
| Outcome: | The proposed methods improve factuality, reasoning, interpretability, and robustness of the generated content. |
The ART of LLM Refinement: Ask, Refine, and Trust (2024.naacl-long)
Copied to clipboard
Kumar Shridhar, Koustuv Sinha, Andrew Cohen, Tianlu Wang, Ping Yu, Ramakanth Pasunuru, Mrinmaya Sachan, Jason Weston, Asli Celikyilmaz
| Challenge: | Large Language Models (LLMs) have demonstrated remarkable generative abilities, but can they judge the quality of their own generations and self-improve? |
| Approach: | They propose a reasoning with a refinement strategy called *ART: Ask, Refine, and Trust* that asks necessary questions to decide when an LLM should refine its output and uses it to affirm or deny trust. |
| Outcome: | The proposed reasoning with a refinement strategy achieves a performance gain of +5 points over baselines on two multistep reasoning tasks. |
Leveraging Abstract Meaning Representation for Knowledge Base Question Answering (2021.findings-acl)
Copied to clipboard
Pavan Kapanipathi, Ibrahim Abdelaziz, Srinivas Ravishankar, Salim Roukos, Alexander Gray, Ramón Fernandez Astudillo, Maria Chang, Cristina Cornelio, Saswati Dana, Achille Fokoue, Dinesh Garg, Alfio Gliozzo, Sairam Gurajada, Hima Karanam, Naweed Khan, Dinesh Khandelwal, Young-Suk Lee, Yunyao Li, Francois Luus, Ndivhuwo Makondo, Nandana Mihindukulasooriya, Tahira Naseem, Sumit Neelam, Lucian Popa, Revanth Gangi Reddy, Ryan Riegel, Gaetano Rossiello, Udit Sharma, G P Shrivatsa Bhargav, Mo Yu
| Challenge: | Existing approaches face challenges including complex question understanding and lack of large end-to-end training datasets. |
| Approach: | They propose a modular knowledge base question answering system that leverages AMR parses for task-independent question understanding. |
| Outcome: | The proposed system achieves state-of-the-art performance on two prominent KBQA datasets based on DBpedia. |
MEENA (PersianMMMU): Multimodal-Multilingual Educational Exams for N-level Assessment (2026.findings-eacl)
Copied to clipboard
Omid Ghahroodi, Arshia Hemmat, Marzia Nouri, Seyed Mohammad Hadi Hosseini, Doratossadat Dastgheib, Mohammad Vali Sanian, Alireza Sahebi, Reihaneh Zohrabi, Mohammad Hossein Rohban, Ehsaneddin Asgari, Mahdieh Soleymani Baghshah
| Challenge: | Recent advances in large vision-language models have primarily focused on English, with limited attention given to other languages. |
| Approach: | They propose a dataset to evaluate Persian VLMs across scientific, reasoning, and human-level understanding tasks. |
| Outcome: | The proposed model performs well across scientific reasoning, reasoning, and human-level understanding tasks in Persian and English. |
Visual–Linguistic Abductive Reasoning with LLMs for Knowledge-based Visual Question Answering (2026.findings-eacl)
Copied to clipboard
| Challenge: | Recent efforts to leverage large language models for reasoning focus on visual perception and language reasoning as separate processes. |
| Approach: | They propose a method that integrates visual and linguistic modalities into interpretable abductive reasoning chains. |
| Outcome: | The proposed method improves performance on AOKVQA, OKVQA and GQA by 2.31% . it uses fuzzy scoring to select the most coherent combination, enabling unified reasoning . |
NOAHQA: Numerical Reasoning with Interpretable Graph Question Answering Dataset (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Existing question answering datasets lack numerical reasoning and reasoning processes . current research on numerical reasoning focuses on simple calculations . |
| Approach: | They propose a conversational and bilingual question answering dataset with numerical reasoning with compound mathematical expressions. |
| Outcome: | The proposed model achieves 55.5 exact match scores while human performance is 89.7. |
Parallelism and Generation Order in Masked Diffusion Language Models: Limits Today, Potential Tomorrow (2026.findings-acl)
Copied to clipboard
Yangyang Zhong, Yanmei Gu, Zhengqing Zang, Xiaomeng Li, Yuqi Ding, Xibei Jia, Yuting Shen, Zhenzhong Lan, Liwang Zhu, Weiping Liu, Junlin Zhou, Haisheng Liu, Zhong Xin Yu, Pengxin Luo, Donglian Qi, Yunfeng Yan, Junbo Zhao
| Challenge: | Autoregressive (AR) language models dominate modern natural language processing due to strong likelihood-based training objectives and reliable left-to-right decoding. |
| Approach: | They characterize MDLM behavior along two dimensions: parallelism strength and generation order . authors propose a Generate-then-Edit paradigm that mitigates dependency loss . |
| Outcome: | The proposed model improves on tasks that require "backward information" the Generate-then-Edit paradigm improves parallel decoding efficiency while reducing dependency loss. |
SANDWiCH: Semantical Analysis of Neighbours for Disambiguating Words in Context ad Hoc (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent studies show that language understanding offered by chat-based Large Language Models is limited and far from human-like performance. |
| Approach: | They propose a framework for multilingual Word Sense Disambiguation using group algebra. |
| Outcome: | The proposed framework surpasses the performance of current alternatives even in low-resource languages while reducing the parameter count by 72%. |
Attribution-Based Analysis and Optimization of Modular Agentic Workflows (2026.findings-acl)
Copied to clipboard
Yingxuan Yang, Bo Huang, Siyuan Qi, Chao Feng, Haoyi Hu, Yuxuan Zhu, Jinbo Hu, Haoran Zhao, Ziyi He, Xiao Liu, ZongYu Wang, Muning Wen, Lin Qiu, Xuezhi Cao, Xunliang Cai, Yong Yu, Weinan Zhang
| Challenge: | Large Language Models (LLMs) have driven the rise of agentic workflows . yet, how can we attribute performance gains to individual upgrades and their interactions? |
| Approach: | They propose a game-theoretic framework that models component upgrades as players and evaluates component coalitions to compute Shapley values. |
| Outcome: | The proposed framework provides interaction-aware attribution and recommendation for model allocation under a fixed workflow structure. |
RAG-Star: Enhancing Deliberative Reasoning with Retrieval Augmented Verification and Refinement (2025.naacl-long)
Copied to clipboard
| Challenge: | Existing large language models (LLMs) show exceptional problem-solving capabilities but struggle with complex reasoning tasks. |
| Approach: | They propose a novel RAG approach that integrates retrieved information to guide tree-based reasoning process based on LLMs. |
| Outcome: | The proposed approach outperforms existing methods in large language models . iteratively plans intermediate sub-queries and answers based on the LLM itself . |
Connecting the Dots: A Knowledgeable Path Generator for Commonsense Question Answering (2020.findings-emnlp)
Copied to clipboard
| Challenge: | Existing QA systems do not have commonsense knowledge or cannot reason with it. |
| Approach: | They propose to augment a general commonsense QA framework with a knowledgeable path generator by extrapolating existing paths from a KG with 'state-of-the-art' language model. |
| Outcome: | The generated paths are interpretable, novel, and relevant to the task. |
JointLK: Joint Reasoning with Language Models and Knowledge Graphs for Commonsense Question Answering (2022.naacl-main)
Copied to clipboard
| Challenge: | Existing KG-augmented models for commonsense question answering ignore the effectively fusing and reasoning over question context representations and the KG representations. |
| Approach: | They propose a novel model which combines a logical reasoning and a dynamic pruning mechanism to solve these limitations. |
| Outcome: | The proposed model improves existing models and performs interpretable reasoning on the CommonsenseQA and OpenBookQA datasets. |
A Regex Minimization Benchmark: A PSPACE-Complete Challenge for Language Models (2026.eacl-long)
Copied to clipboard
| Challenge: | Language models (LMs) have demonstrated impressive reasoning capabilities across domains . but their ability to handle PSPACE-complete problems remains underexplored . a new benchmark for regex minimization is proposed to evaluate LMs' reasoning capabilities . |
| Approach: | They propose a benchmark for regex minimization to evaluate LMs' reasoning power . they use a million regexes paired with their minimal equivalents to evaluate their performance . |
| Outcome: | The proposed model can solve NP-complete problems, but their ability to handle PSPACE-complete ones remains underexplored. |
The Reasoning Trap: How Enhancing LLM Reasoning Amplifies Tool Hallucination (2026.acl-long)
Copied to clipboard
| Challenge: | Recent studies suggest that strengthening reasoning often coincides with increased hallucination . however, no prior work has examined whether reasoning enhancement itself causes tool hallucinism . |
| Approach: | They propose a diagnostic benchmark measuring tool hallucination in two failure modes . they demonstrate a causal relationship between enhancing reasoning and tool hallubulation . |
| Outcome: | The proposed benchmark measures tool hallucination in two failure modes: no tool available, and (ii) only distractor tools available. |
Verifiable, Debuggable, and Repairable Commonsense Logical Reasoning via LLM-based Theory Resolution (2024.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models (LLMs) have led to substantial interest in their application to commonsense reasoning tasks. |
| Approach: | They propose a logical reasoning framework that integrates commonsense knowledge with a verifiable logical framework that mitigates hallucinations and facilitates debugging. |
| Outcome: | The proposed framework improves on three language-based reasoning tasks and improves accuracy and reasoning correctness. |
Look Light, Think Heavy: What Multimodal Chain-of-Thought Reasoning Can and Cannot Do (2026.acl-long)
Copied to clipboard
| 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. |
Subgraph Retrieval Enhanced Model for Multi-hop Knowledge Base Question Answering (2022.acl-long)
Copied to clipboard
| Challenge: | Existing retrieval methods for knowledge base question answering are either heuristic or interwoven with the reasoning, causing reasoning on the partial subgraphs. |
| Approach: | They propose a subgraph retrieval framework that decouples the retrieval from the subsequent reasoning process and trains subgraphs for easier reasoning. |
| Outcome: | The proposed framework improves retrieval and QA performance over existing methods. |
Revealing the Parallel Multilingual Learning within Large Language Models (2024.emnlp-main)
Copied to clipboard
Yongyu Mu, Peinan Feng, Zhiquan Cao, Yuzhang Wu, Bei Li, Chenglong Wang, Tong Xiao, Kai Song, Tongran Liu, Chunliang Zhang, JingBo Zhu
| Challenge: | Large language models (LLMs) can handle multilingual and cross-lingual text within a single input; however, previous studies focusing on using English as the pivot language to enhance language understanding and reasoning focus on using multiple languages. |
| Approach: | They propose to use parallel multilingual input to enhance the model's comprehension of the input and to examine how multilingual processing affects prediction. |
| Outcome: | The proposed model can handle multilingual and cross-lingual text within a single input, but previous studies focused on using English as the pivot language to enhance language understanding and reasoning. |
ReFT: Reasoning with Reinforced Fine-Tuning (2024.acl-long)
Copied to clipboard
| Challenge: | Existing approaches to improve the generalization of large language models are using Supervised Fine-Tuning (SFT) this approach does not show sufficient generalization ability because it only relies on the given CoT data. |
| Approach: | They propose to use Chain-of-Thought annotations to train Large Language Models using supervised fine-tuning to improve generalization. |
| Outcome: | The proposed approach outperforms SFT on GSM8K, MathQA, and SVAMP datasets and shows a superior generalization ability. |
TurkishMMLU: Measuring Massive Multitask Language Understanding in Turkish (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing multiple choice question answering benchmarks employ automatic translation for multilingual evaluation, but this approach is error-prone and potentially introduces culturally biased questions. |
| Approach: | They introduce the first multitask, multiple-choice Turkish QA benchmark, TurkishMMLU . they evaluate over 20 LLMs including open-source, closed-source and Turkish-adapted models . |
| Outcome: | The proposed benchmarks evaluate the reasoning, comprehension, and mathematical abilities of large language models. |
Thinking Like a Skeptic: Defeasible Inference in Natural Language (2020.findings-emnlp)
Copied to clipboard
Rachel Rudinger, Vered Shwartz, Jena D. Hwang, Chandra Bhagavatula, Maxwell Forbes, Ronan Le Bras, Noah A. Smith, Yejin Choi
| Challenge: | Defeasible inference is a mode of reasoning in which an inference may be weakened or overturned in light of new evidence. |
| Approach: | They propose a dataset for defeasible inference in natural language that includes extensions to existing inference datasets. |
| Outcome: | Defeasible NLI extends existing datasets for defeaasibility inference in natural language . generative models can weaken or strengthen inferences up to 68% of the time, it shows . |
Tiered Reasoning for Intuitive Physics: Toward Verifiable Commonsense Language Understanding (2021.findings-emnlp)
Copied to clipboard
| Challenge: | Large-scale, pre-trained language models (LMs) have achieved human-level performance on a breadth of language understanding tasks. |
| Approach: | They propose a commonsense reasoning dataset with dense annotations that allows multi-tiered evaluation of machines’ reasoning process. |
| Outcome: | The proposed model can achieve high end performance but struggle to support predictions with valid supporting evidence. |
Deceptive Semantic Shortcuts on Reasoning Chains: How Far Can Models Go without Hallucination? (2024.naacl-long)
Copied to clipboard
| Challenge: | Existing large language models (LLMs) suffer from hallucinations and unfaithful reasoning due to keyword/entity biases. |
| Approach: | They propose a new probing method and benchmark to quantify this phenomenon by using a keyword/entity biases-based probing technique called EUREQA. |
| Outcome: | The proposed method achieves 62% accuracy on multi-hop and complex QA benchmarks. |
Retrieval and Reasoning on KGs: Integrate Knowledge Graphs into Large Language Models for Complex Question Answering (2024.findings-emnlp)
Copied to clipboard
| 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. |
Teaching LLMs to Plan, Not Just Solve: Plan Learning Boosts LLMs Generalization in Reasoning Tasks (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for reinforcement learning (RL) on self-generated data are limited in many domains. |
| Approach: | a new framework combines plan-based search with Step-level Advantage Preference Optimization to optimize plan learning. |
| Outcome: | The proposed framework improves in-domain performance and out-of-domain benchmarks. |
RRHF-V: Ranking Responses to Mitigate Hallucinations in Multimodal Large Language Models with Human Feedback (2025.coling-main)
Copied to clipboard
| Challenge: | Existing methods to mitigate hallucinations generate erroneous or fabricated information. |
| Approach: | They propose a rank-response-based model that annotates pair-reponses and trains alignment algorithms to improve the correspondence between images and text. |
| Outcome: | The proposed model outperforms the DPO method and outperfies existing methods on two MLLMs of different sizes and four widely used benchmarks. |
ReasoningRec: Bridging Personalized Recommendations and Human-Interpretable Explanations through LLM Reasoning (2025.findings-naacl)
Copied to clipboard
| Challenge: | Empirical evaluations demonstrate that ReasoningRec surpasses state-of-the-art methods by up to 12.5% in recommendation prediction while simultaneously providing human-intelligible explanations. |
| Approach: | They propose a reasoning-based recommendation framework that leverages Large Language Models to model users and items, focusing on preferences, aversions, and explanatory reasoning. |
| Outcome: | The proposed framework surpasses state-of-the-art methods by up to 12.5% in recommendation prediction while providing human-intelligible explanations. |
From Heuristic to Analytic: Cognitively Motivated Strategies for Coherent Physical Commonsense Reasoning (2023.emnlp-main)
Copied to clipboard
| Challenge: | Pre-trained language models have shown impressive performance in various language tasks, but are prone to spurious correlations and illusory information. |
| Approach: | They propose to use pre-trained language models to justify decisions with formalized, coherent reasoning chains. |
| Outcome: | The proposed strategies improve coherence of rationalizations yielding state-of-the-art results on Tiered Reasoning for Intuitive Physics (TRIP). |
Could you give me a hint ? Generating inference graphs for defeasible reasoning (2021.findings-acl)
Copied to clipboard
| Challenge: | Defeasible reasoning is a mode of reasoning where conclusions can be overturned by taking into account new evidence. |
| Approach: | They propose to automatically generate inference graphs for a defeasible inference task by transfer learning from a related NLP task. |
| Outcome: | The proposed method generates meaningful graphs for a defeasible inference task and human accuracy improves by 20%. |
PLAWBENCH: A Rubric-Based Benchmark for Evaluating LLMs in Real-World Legal Practice (2026.acl-long)
Copied to clipboard
Yuzhen Shi, Huanghai Liu, Yiran HU, Song Gaojie, Xu Xinran, Yubo Ma, Tianyi Tang, Li Zhang, Qingjing Chen, Feng Di, Wenbo Lv, Weiheng Wu, Kexin Yang, Sen Yang, Wei Wang, Rongyao Shi, Qiu Yuanyang, Yuemeng Qi, Zhang Jingwen, Sui Xiaoyu, Yifan Chen, Zhang Yi, An Yang, Bowen Yu, Dayiheng Liu, Junyang Lin, Weixing Shen, Bing Zhao, Charles L. A. Clarke, HU Wei
| Challenge: | Existing benchmarks for large language models (LLMs) are coarse, single-dimensional metrics and do not explicitly assess fine-grained legal reasoning. |
| Approach: | They propose a Practical Law Benchmark to evaluate large language models in real-world legal practice scenarios. |
| Outcome: | The proposed model is based on 850 questions and 13 scenarios with expert-designed evaluation rubrics. |
FastCuRL: Curriculum Reinforcement Learning with Stage-wise Context Scaling for Efficient Training R1-like Reasoning Models (2025.findings-emnlp)
Copied to clipboard
| 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. |
Argumentation Mining on Essays at Multi Scales (2020.coling-main)
Copied to clipboard
| Challenge: | Argumentation mining on essays is a new task in natural language processing. |
| Approach: | They propose a multi-scale argumentation mining model which aims to identify the types and locations of argumentation components from essay text. |
| Outcome: | The proposed model outperforms existing models on mining all types of argumentation components on the Persuasive Essay dataset. |
bgGLUE: A Bulgarian General Language Understanding Evaluation Benchmark (2023.acl-long)
Copied to clipboard
Momchil Hardalov, Pepa Atanasova, Todor Mihaylov, Galia Angelova, Kiril Simov, Petya Osenova, Veselin Stoyanov, Ivan Koychev, Preslav Nakov, Dragomir Radev
| Challenge: | bgGLUE is a benchmark for evaluating language models on natural language understanding (NLU) tasks in Bulgarian. |
| Approach: | They propose to use a benchmark to evaluate language models on NLU tasks in Bulgarian. |
| Outcome: | The proposed model performs well on sequence labeling tasks, but there is room for improvement for tasks that require more complex reasoning. |
Explaining Black Box Predictions and Unveiling Data Artifacts through Influence Functions (2020.acl-main)
Copied to clipboard
| Challenge: | Modern deep learning models for NLP are notoriously opaque, and this has motivated efforts to design example-specific approaches to interpret such models. |
| Approach: | They propose to use influence functions to explain models by highlighting important words in input text to provide models with an explanation. |
| Outcome: | The proposed approach is particularly useful for natural language inference, a task in which ‘saliency maps’ may not have clear interpretation. |
ReasonBERT: Pre-trained to Reason with Distant Supervision (2021.emnlp-main)
Copied to clipboard
| Challenge: | Existing pre-training methods only harvest learning signals from local contexts of naturally occurring texts . ReasonBert provides a method for reasoning over long-range relations and multiple, possibly hybrid contexts. |
| Approach: | They propose a method that augments language models with the ability to reason over long-range relations and multiple, possibly hybrid contexts. |
| Outcome: | The proposed method significantly improves sample efficiency over strong baselines. |
TRACE the Evidence: Constructing Knowledge-Grounded Reasoning Chains for Retrieval-Augmented Generation (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing retrievers are not perfect and often include irrelevant documents in the retrieved set. |
| Approach: | They propose to construct knowledge-grounded reasoning chains from retrieved documents to integrate supporting evidence into RAG models. |
| Outcome: | The proposed model achieves an average performance improvement of 14.03% on three multi-hop QA datasets. |
Reasoning with Language Model is Planning with World Model (2023.emnlp-main)
Copied to clipboard
| 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. |
Bi-Chainer: Automated Large Language Models Reasoning with Bidirectional Chaining (2024.findings-acl)
Copied to clipboard
| Challenge: | Existing unidirectional chaining methods suffer from low prediction accuracy and efficiency. |
| Approach: | They propose a bidirectional chaining method which dynamically switches to depth-first reasoning in the opposite reasoning direction when it encounters multiple branching options within the current direction. |
| Outcome: | The proposed method achieves sizable accuracy boots over unidirectional chaining frameworks on four challenging logical reasoning datasets. |
CofiPara: A Coarse-to-fine Paradigm for Multimodal Sarcasm Target Identification with Large Multimodal Models (2024.acl-long)
Copied to clipboard
| Challenge: | Current methods for multimodal sarcasm target identification focus on superficial indicators in an end-to-end manner, overlooking the nuanced understanding of multimodal content. |
| Approach: | They propose a multimodal sarcasm target identification framework with a coarse-to-fine paradigm by augmenting sarcasm explainability with reasoning and pre-training knowledge. |
| Outcome: | The proposed framework outperforms state-of-the-art methods and exhibits explainability in deciphering sarcasm as well. |
Document-Level Event Argument Extraction With a Chain Reasoning Paradigm (2023.acl-long)
Copied to clipboard
| Challenge: | Document-level event argument extraction aims to identify event arguments beyond sentence level, where a significant challenge is to model long-range dependencies. |
| Approach: | They propose a chain reasoning paradigm which captures long-range interdependence due to the chains’ compositional nature and generates decomposable first-order logic rules for reasoning. |
| Outcome: | The proposed method outperforms previous methods on two benchmarks and is robust enough to defend against adversarial attacks. |
Meta-rater: A Multi-dimensional Data Selection Method for Pre-training Language Models (2025.acl-long)
Copied to clipboard
Xinlin Zhuang, Jiahui Peng, Ren Ma, Yinfan Wang, Tianyi Bai, Xingjian Wei, Qiu Jiantao, Chi Zhang, Ying Qian, Conghui He
| Challenge: | composition of pre-training datasets for large language models remains undisclosed . current methods for evaluating data quality are limited by single-dimensional evaluation or redundancy-focused strategies. |
| Approach: | They propose a multi-dimensional data selection method that integrates dimensions with existing quality metrics through learned optimal weightings. |
| Outcome: | The proposed method doubles convergence speed for 1.3B model models and improves downstream task performance by 3.23%. |
LiveFact: A Dynamic, Time-Aware Benchmark for LLM-Driven Fake News Detection (2026.acl-long)
Copied to clipboard
| Challenge: | Current evaluation frameworks are static and vulnerable to benchmark data contamination . current models are ineffective at assessing reasoning under temporal uncertainty . |
| Approach: | They propose a live-based benchmark that simulates the real-world "fog of war" they propose evaluating models on their ability to reason with evolving, incomplete information . |
| Outcome: | The proposed model outperforms proprietary state-of-the-art models in classification and evidence mode . it also provides a component to monitor BDC explicitly . |
Accelerating LLM Reasoning via Early Rejection with Partial Reward Modeling (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) are increasingly relied upon for solving complex reasoning tasks. |
| Approach: | They propose to use Process Reward Models to scale inference time compute by generating in parallel . they propose to provide early signals that enable the rejection of suboptimal candidates before full generation of step is complete. |
| Outcome: | The proposed method achieves 1.4 – 9 reduction in inference FLOPs without degrading final performance. |
OpenGenAlign: A Preference Dataset and Benchmark for Trustworthy Reward Modeling in Open-Ended, Long-Context Generation (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing reward models perform suboptimal on held-out benchmarks, resulting in poor quality outputs. |
| Approach: | They propose a framework and a high-quality dataset to evaluate reward models . they define four key metrics to assess generation quality and develop a pipeline to evaluate outputs . |
| Outcome: | The proposed framework and dataset improves hallucination-free, comprehensive, reliable, and efficient open-ended long-context generation. |
NatLogAttack: A Framework for Attacking Natural Language Inference Models with Natural Logic (2023.acl-long)
Copied to clipboard
| Challenge: | Despite the recent advances in distributed representation and neural networks, it remains an open question whether the models perform real reasoning to reach their conclusions or rely on spurious correlations. |
| Approach: | They propose to use logic formalism to perform systematic attacks centring around natural logic to generate better adversarial examples with fewer visits to the victim models. |
| Outcome: | The proposed framework generates better adversarial examples with fewer visits to the victim models. |
TC–RAG: Turing–Complete RAG’s Case study on Medical LLM Systems (2025.acl-long)
Copied to clipboard
Xinke Jiang, Yue Fang, Rihong Qiu, Haoyu Zhang, Yongxin Xu, Hao Chen, Wentao Zhang, Ruizhe Zhang, Yuchen Fang, Xinyu Ma, Xu Chu, Junfeng Zhao, Yasha Wang
| Challenge: | Existing approaches to RAG neglect system state variables, resulting in poor performance and erroneous knowledge accumulation. |
| Approach: | They propose a framework that incorporates a Turing Complete System to manage state variables and manage retrieval halting. |
| Outcome: | The proposed framework improves on seven real-world healthcare datasets and shows that it is more accurate than existing methods. |
Reasoning Gets Harder for LLMs Inside A Dialogue (2026.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) achieve strong performance on many reasoning benchmarks, yet these evaluations typically focus on isolated tasks that differ from real-world usage in task-oriented dialogue (TOD). |
| Approach: | They propose to use a dynamic benchmark to examine how framing reasoning tasks within task-oriented dialogue (TOD) affect LLM performance. |
| Outcome: | The proposed model performs well on isolated tasks and in task-oriented dialogues, but performance is inconsistent between them. |
AgentThink: A Unified Framework for Tool-Augmented Chain-of-Thought Reasoning in Vision-Language Models for Autonomous Driving (2025.findings-emnlp)
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
| 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% . |
Navigating Hallucinations for Reasoning of Unintentional Activities (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Existing models of intentionality recognition struggle to understand the reasoning behind unintentional actions. |
| Approach: | They propose a novel prompting technique which allows the model to navigate through hallucinated thoughts to achieve better reasoning. |
| Outcome: | The proposed prompting technique outperforms standard prompting while minimizing hallucinations. |
Sarcasm-R1: Enhancing Sarcasm Detection through Focused Reasoning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for sarcasm detection are limited by supervised learning or prompt engineering . a new approach decomposes sarcasm detection into three dimensions: language, context, and emotion . |
| Approach: | They propose a method that decomposes sarcasm detection into three dimensions: language, context, and emotion. |
| Outcome: | The proposed method outperforms state-of-the-art methods in most cases. |
MetaScale: Test-Time Scaling with Evolving Meta-Thoughts (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches impose fixed cognitive structures that enhance performance in specific tasks but lack adaptability across diverse scenarios. |
| Approach: | They propose a test-time scaling framework based on meta-thoughts to improve performance . meta-thinkts are adaptive thinking strategies tailored to a given task . |
| Outcome: | Experimental results show that MetaScale outperforms standard inference approaches . it can scale more effectively with increasing sampling budgets and produces more structured responses . |
Multilingual Reasoning via Self-training (2025.naacl-long)
Copied to clipboard
| Challenge: | Recent studies have introduced eclectic strategies to improve reasoning beyond English, but these methods are related to specific language that is not always optimal for reasoning. |
| Approach: | They propose a modular approach that instructs models to structure reasoning passages in a different problem space and then self-refines their capabilities to deliver step-wise reasoning passage. |
| Outcome: | The proposed approach achieves significant improvements in multilingual reasoning of various models and task, with improved reasoning consistency across languages. |
Understanding Faithfulness and Reasoning of Large Language Models on Plain Biomedical Summaries (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Generating plain biomedical summaries with Large Language Models (LLMs) can enhance access to biomedically knowledge. |
| Approach: | They propose a benchmark dataset with expert-annotated Faithfulness and Reasoning on plain biomedical summaries. |
| Outcome: | The proposed dataset shows that LLMs perform poorly in generating faithful biomedical summaries and that abstractiveness and faithfulness are negatively correlated. |
BERT Knows Punta Cana is not just beautiful, it’s gorgeous: Ranking Scalar Adjectives with Contextualised Representations (2020.emnlp-main)
Copied to clipboard
| Challenge: | Adjectives describe positive properties of nouns but with different intensity. |
| Approach: | They propose a BERT-based approach to intensity detection for scalar adjectives by generating vectors directly from contextualised representations. |
| Outcome: | The proposed model outperforms static embeddings and previous models with dedicated resources on an Indirect Question Answering task. |
Adapt to Thrive! Adaptive Power-Mean Policy Optimization for Improved LLM Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for reinforcement learning with verifiable rewards (RLVR) rely on static objective functions and rigid clipping strategies that misalign with the model’s evolving reasoning capabilities. |
| Approach: | They propose to incorporate Power-Mean Policy Optimization (PMPO) and Feedback-Adaptive Clipping (FAC) to overcome limitations of static mechanisms. |
| Outcome: | Extensive experiments on nine reasoning tasks show the proposed paradigm outperforms state-of-the-art methods. |
Agentic-R1: Distilled Dual-Strategy Reasoning (2025.emnlp-main)
Copied to clipboard
| Challenge: | Current long chain-of-thought models rely on slow and error-prone natural language traces. |
| Approach: | They propose a framework that distills complementary reasoning strategies from multiple teachers into a unified student model. |
| Outcome: | The proposed framework improves accuracy on computation-intensive tasks and reduces inference latency on standard benchmarks. |
Beneath the Surface: Unveiling Harmful Memes with Multimodal Reasoning Distilled from Large Language Models (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing methods for harmful meme detection ignore in-depth cognition of meme text and image . authors propose a framework for learning reasonable thoughts from LLMs for better multimodal fusion . |
| Approach: | They propose to use large language models to learn reasonable thoughts from LLMs for better multimodal fusion and lightweight fine-tuning. |
| Outcome: | The proposed approach achieves superior performance than state-of-the-art methods on the harmful meme detection task. |
MCˆ2: Multi-perspective Convolutional Cube for Conversational Machine Reading Comprehension (P19-1)
Copied to clipboard
| Challenge: | Existing models combine previous questions for conversation understanding and only employ recurrent neural networks (RNN) for reasoning. |
| Approach: | They propose a multi-perspective convolutional cube model that integrates 1D and 2D convolutions with recurrent neural networks (RNN) to understand context from different perspectives. |
| Outcome: | The proposed model is based on the Conversational Question Answering (CoQA) dataset and achieves state-of-the-art results. |
Unraveling Misinformation Propagation in LLM Reasoning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have demonstrated impressive capabilities in reasoning, but how they propagate within their reasoning process remains underexplored. |
| Approach: | They propose a practical approach to mitigating misinformation propagation in LLMs by applying factual corrections early in the reasoning process and fine-tuning on synthesized data with early-stage corrections significantly improves reasoning factuality. |
| Outcome: | The proposed model can correct misinformation when explicitly instructed, but fails to correct misinformation less than half the time even with explicit instructions. |
CLaw: Benchmarking Chinese Legal Knowledge in Large Language Models - A Fine-grained Corpus and Reasoning Analysis (2025.findings-emnlp)
Copied to clipboard
| Challenge: | a new benchmark is designed to evaluate LLMs on Chinese legal knowledge and its application in reasoning . general pre-training that ingests legal texts without specialized focus compromises reliability of LLM responses . achieving trustworthy legal reasoning in LLM requires a robust synergy of accurate knowledge retrieval and strong general reasoning capabilities. |
| Approach: | They propose a benchmark specifically engineered to evaluate LLMs on Chinese legal knowledge and its application in reasoning. |
| Outcome: | The proposed benchmark evaluates LLMs on Chinese legal knowledge and its application in reasoning. |
Disentangled Multi-span Evolutionary Network against Temporal Knowledge Graph Reasoning (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing methods for temporal knowledge Graphs neglect internal structural interactions between subgraphs and ignore potential smooth features that do not lead to semantic changes. |
| Approach: | They propose to use a disentangled multi-span evolutionary network to capture local neighbor features while perceiving historical neighbor semantic information. |
| Outcome: | Extensive experiments show that the proposed model outperforms the state-of-the-art in TKG reasoning by 22.7%. |
LightThinker: Thinking Step-by-Step Compression (2025.emnlp-main)
Copied to clipboard
Jintian Zhang, Yuqi Zhu, Mengshu Sun, Yujie Luo, Shuofei Qiao, Lun Du, Da Zheng, Huajun Chen, Ningyu Zhang
| Challenge: | Recent advances in Large Language Models have demonstrated their remarkable capabilities in complex reasoning tasks, but their efficiency is hindered by the substantial memory and computational costs associated with generating lengthy tokens. |
| Approach: | They propose a method that dynamically compresses verbose thought steps into compact representations and discards original reasoning chains. |
| Outcome: | The proposed method reduces peak memory usage and inference time while maintaining competitive accuracy. |
MINED: Probing and Updating with Multimodal Time-Sensitive Knowledge for Large Multimodal Models (2026.findings-acl)
Copied to clipboard
Kailin Jiang, Ning Jiang, Yuntao Du, Yuchen Ren, Yuchen Li, Yifan Gao, Jinhe Bi, Yunpu Ma, Bin Li, Lei Liu, Qing Li
| Challenge: | Existing benchmarks for Large Multimodal Models (LMMs) are constrained by static representations, inadequately evaluating their ability to understand time-sensitive knowledge. |
| Approach: | They propose a benchmark containing 2,104 time-sensitive knowledge samples spanning six knowledge types to evaluate temporal awareness along 6 key dimensions and 11 challenging tasks. |
| Outcome: | The proposed benchmark measures temporal awareness along 6 key dimensions and 11 tasks, while most open-source LMMs still lack time understanding ability. |
RoseRAG: Robust Retrieval-augmented Generation with Small-scale LLMs via Margin-aware Preference Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) have impressive performance but face high computational costs and latency, limiting their deployment in resource-constrained settings. |
| Approach: | They propose a robust RAG framework for large language models via Margin-aware Preference Optimization to enhance the accuracy and reliability of SLMs. |
| Outcome: | The proposed framework surpasses state-of-the-art benchmarks on three open-domain question answering tasks. |
Bridging the Memorization-Utilization Gap: Near-Lossless Context Compression via Reinforcement Learning (2026.acl-long)
Copied to clipboard
| Challenge: | Recent advances in context compression have failed to effectively utilize compressed representations for downstream tasks. |
| Approach: | They propose a holistic training paradigm that uses outcome-based RL to enable implicit expansion. |
| Outcome: | The proposed model outperforms previous models on NIAH, LongBench and multi-hop reasoning. |
What’s the Meaning of Superhuman Performance in Today’s NLU? (2023.acl-long)
Copied to clipboard
Simone Tedeschi, Johan Bos, Thierry Declerck, Jan Hajič, Daniel Hershcovich, Eduard Hovy, Alexander Koller, Simon Krek, Steven Schockaert, Rico Sennrich, Ekaterina Shutova, Roberto Navigli
| Challenge: | Recent research has focused on developing larger pretrained language models and introducing benchmarks such as SuperGLUE and SQuAD to measure their abilities. |
| Approach: | They propose to use benchmarks such as SuperGLUE and SQUAD to evaluate PLMs' abilities in language understanding, reasoning, and reading comprehension to assess their performance. |
| Outcome: | The proposed benchmarks have serious limitations affecting comparison between humans and PLMs and provide recommendations for fairer and more transparent benchmarks. |
Neutralizing Bias in LLM Reasoning using Entailment Graphs (2025.findings-acl)
Copied to clipboard
| Challenge: | Natural Language Inference (NLI) is a foundational understanding task in language understanding. |
| Approach: | They propose a framework to construct counterfactual reasoning data and fine-tune LLMs to reduce attestation bias. |
| Outcome: | The proposed framework reduces hallucinations from attestation bias on original and bias-neutralized datasets while keeping hypotheses unchanged. |
TransferTOD: A Generalizable Chinese Multi-Domain Task-Oriented Dialogue System with Transfer Capabilities (2024.emnlp-main)
Copied to clipboard
Ming Zhang, Caishuang Huang, Yilong Wu, Shichun Liu, Huiyuan Zheng, Yurui Dong, Yujiong Shen, Shihan Dou, Jun Zhao, Junjie Ye, Qi Zhang, Tao Gui, Xuanjing Huang
| Challenge: | Current datasets cater to user-led systems and are limited to predefined specific scenarios and slots. |
| Approach: | They propose to use a Chinese dialogue dataset to train a model that authentically simulates human-computer dialogues in 30 popular life service scenarios. |
| Outcome: | The proposed model achieves a joint accuracy of 75.09% in out-of-domain evaluations . it also achieves notable abilities in slot filling and questioning . |
A Collaborative Reasoning Framework Powered by Reinforcement Learning and Large Language Models for Complex Questions Answering over Knowledge Graph (2025.coling-main)
Copied to clipboard
| Challenge: | Knowledge Graph Question Answering (KGQA) aims to answer natural language questions by reasoning across multiple triples in knowledge graphs. |
| Approach: | They propose a collaborative reasoning framework powered by RL and LLMs to answer complex questions based on the knowledge graph. |
| Outcome: | The proposed model surpasses state-of-the-art models on four datasets. |
From Spatial Relations to Spatial Configurations (2020.lrec-1)
Copied to clipboard
| Challenge: | Existing spatial representations are not sufficient for describing complex spatial configurations. |
| Approach: | They propose to integrate existing spatial representation languages with an annotation schema to extend the capabilities of existing ones. |
| Outcome: | The proposed language can represent a large set of spatial concepts crucial for reasoning . it integrates with the Abstract Meaning Representation (AMR) annotation schema and annotates text from diverse datasets . |
ACQUIRED: A Dataset for Answering Counterfactual Questions In Real-Life Videos (2023.emnlp-main)
Copied to clipboard
Te-Lin Wu, Zi-Yi Dou, Qingyuan Hu, Yu Hou, Nischal Chandra, Marjorie Freedman, Ralph Weischedel, Nanyun Peng
| Challenge: | despite its importance, there are few datasets that cover multimodal counterfactual reasoning . a dataset focusing on this area is limited because of its limited coverage over synthetic environments . |
| Approach: | They develop a video question answering dataset that provides questions on multimodal reasoning . they ask questions about counterfactual hypotheses over visual events . |
| Outcome: | The proposed dataset shows a significant performance gap between models and humans . it provides questions that span physical, social, and temporal dimensions . |
A Closer Look at Bias and Chain-of-Thought Faithfulness of Large (Vision) Language Models (2025.findings-emnlp)
Copied to clipboard
| 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. |
How Susceptible Are LLMs to Logical Fallacies? (2024.lrec-main)
Copied to clipboard
| Challenge: | Recent studies have focused on LLMs' reasoning abilities, but their rational thinking capacity is not as robust as that of other NLP downstream tasks. |
| Approach: | They propose a diagnostic benchmark to assess the robustness of Large Language Models against logical fallacies by comparing their performance against a scenario where the persuader employs logical fallsacie. |
| Outcome: | The proposed benchmark compares the performance of LLMs in debates on controversial topics against logical fallacies. |
Revisiting Chain-of-Thought Prompting: Zero-shot Can Be Stronger than Few-shot (2025.findings-emnlp)
Copied to clipboard
| 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. |
MULFE: A Multi-Level Benchmark for Free Text Model Editing (2024.acl-long)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have impressive capabilities in comprehending human language and vast parametric knowledge obtained from large corpora. |
| Approach: | They propose a multi-level benchmark for free text model editing to bridge the gap . they categorize probe queries into three levels of generalization . |
| Outcome: | The proposed method improves the generalization performance of large langugae models. |
When Internalization Fails: Finding Better Targets for Reasoning Compression (2026.findings-acl)
Copied to clipboard
| Challenge: | Reasoning language models generate long reasoning traces that increase latency and cost. |
| Approach: | They compare three approaches to shorten reasoning traces by inference-time truncation . they use Implicit Chain-of-Thought-style curricula that progressively shorten the teacher trace . |
| Outcome: | The proposed methods work well on GSM8K and multiplication tasks. |
Towards Medical Complex Reasoning with LLMs through Medical Verifiable Problems (2025.findings-acl)
Copied to clipboard
| Challenge: | OpenAI o1 has been a significant milestone in large language model development . however, most research in reasoning has focused on mathematical tasks . medical domains require robust reasoning to provide reliable answers . |
| Approach: | They propose a method to verify medical reasoning using a medical verifier . they also propose RL and reinforcement learning to enhance reasoning . |
| Outcome: | The proposed method outperforms general and medical-specific baselines using only 40K verifiable problems. |
TabularMath: Understanding Math Reasoning over Tables with Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Mathematical reasoning has long been a key benchmark for evaluating large language models. |
| Approach: | They propose a framework that transforms math word problems into scalable tabular reasoning tasks. |
| Outcome: | The proposed framework transforms math word problems into scalable and verified tabular reasoning tasks. |
A Frustratingly Easy Plug-and-Play Detection-and-Reasoning Module for Chinese Spelling Check (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Recent years, Chinese Spelling Check (CSC) has been greatly improved by designing task-specific pre-training methods or introducing auxiliary tasks. |
| Approach: | They propose to decompose Chinese Spelling Check into detection, reasoning, and searching subtasks and to train a module that is compatible with existing CSC models. |
| Outcome: | The proposed module can be trained for one model and benefit other models. |
PGPO: Enhancing Agent Reasoning via Pseudocode-style Planning Guided Preference Optimization (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing LLM agents generate verbose and inefficient natural language plans to guide reasoning, which restricts agents’ ability to generalize across similar tasks. |
| Approach: | They propose a pseudocode-style planning guide optimization method that captures the structural logic of reasoning and uses two planning-oriented rewards to enhance agent learning. |
| Outcome: | The proposed method outperforms existing LLM agents on representative agent benchmarks and outperformed the current leading baselines. |
Complex Numerical Reasoning with Numerical Semantic Pre-training Framework (2025.emnlp-main)
Copied to clipboard
| Challenge: | Numerical knowledge graphs (NKGs) are not limited to discrete entity-relation knowledge. |
| Approach: | They propose to combine numerical values and entities to solve multi-hop complex reasoning over incomplete knowledge graphs. |
| Outcome: | The proposed approach handles up to 102 types of complex numerical reasoning queries on three public datasets. |
Prompt-R1: Collaborative Automatic Prompting Framework via End-to-end Reinforcement Learning (2026.findings-acl)
Copied to clipboard
Wenjin Liu, Haoran Luo, Xueyuan Lin, Haoming Liu, Tiesunlong Shen, Jiapu Wang, Rui Mao, Erik Cambria
| Challenge: | Existing large language models are limited in understanding, reasoning, calculation, and generation, limiting their performance in complex reasoning and dynamic tasks. |
| Approach: | They propose a plug-and-play framework that integrates a small-scale LLM (as agent) with large-scale large-level LLMs (a as environment) they propose generating prompts that are used to interact with LLM, and a double constraint reward that optimizes correctness and quality of generation. |
| Outcome: | The proposed framework significantly outperforms baseline large-scale large-language models across various tasks. |
MMBoundary: Advancing MLLM Knowledge Boundary Awareness through Reasoning Step Confidence Calibration (2025.acl-long)
Copied to clipboard
| Challenge: | Existing methods calibrate model confidence on entire response, which leads to incorrect answers with high confidence. |
| Approach: | They propose a framework that advances the knowledge boundary awareness of multimodal large language models through reasoning step confidence calibration. |
| Outcome: | Empirical results show that the proposed framework outperforms existing methods across domains and metrics. |
DC-Instruct: An Effective Framework for Generative Multi-intent Spoken Language Understanding (2024.emnlp-main)
Copied to clipboard
| Challenge: | Existing prompt learning frameworks lack explicit modeling of dual-task dependencies and oversight of task-specific semantic differences among utterances. |
| Approach: | They propose a generative framework based on Dual-task Inter-dependent Instructions (DII) and Supervised Contrastive Instructions that leverages utterance semantics differences by guiding LLMs to determine whether a pair of utterrances share the same or similar labels. |
| Outcome: | The proposed framework outperforms existing models and state-of-the-art methods on public benchmark datasets and shows that it improves SLU reasoning. |
What Makes it Ok to Set a Fire? Iterative Self-distillation of Contexts and Rationales for Disambiguating Defeasible Social and Moral Situations (2023.findings-emnlp)
Copied to clipboard
Kavel Rao, Liwei Jiang, Valentina Pyatkin, Yuling Gu, Niket Tandon, Nouha Dziri, Faeze Brahman, Yejin Choi
| Challenge: | Moral or ethical judgments rely heavily on the contexts in which they occur . a student model that produces defeasible contexts with improved validity, diversity, and defasibility is superior to intermediate student models . |
| Approach: | a new study uses a student model to provide contextualizations that make an action morally acceptable . the model is based on a dataset of 115K defeasible moral actions rated highly by human annotators . |
| Outcome: | The proposed model outperforms all intermediate models in a high-quality dataset . the model is based on 1.2M entries of contextualizations and rationales for 115K moral actions . |
LCR-RAG: Enhancing Logical Consistency in Retrieval-Augmented Generation via Neuro-symbolic Reinforcement Learning (2026.acl-long)
Copied to clipboard
| Challenge: | Retrieval-Augmented Generation (RAG) is widely used to ground large language models in external knowledge and improve factual accuracy. |
| Approach: | They propose a framework that integrates neuro-symbolic verification with reinforcement learning to optimize logical consistency. |
| Outcome: | The proposed framework outperforms strong RAG baselines on hotpotQA, ASQA, and TriviaQA. |
Towards Better Understanding of Program-of-Thought Reasoning in Cross-Lingual and Multilingual Environments (2025.findings-acl)
Copied to clipboard
Patomporn Payoungkhamdee, Pume Tuchinda, Jinheon Baek, Samuel Cahyawijaya, Can Udomcharoenchaikit, Potsawee Manakul, Peerat Limkonchotiwat, Ekapol Chuangsuwanich, Sarana Nutanong
| 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. |
Agent-GWO: Collaborative Agents for Dynamic Prompt Optimization in Large Language Models (2026.findings-acl)
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
| 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. |
Not All Languages Are Created Equal in LLMs: Improving Multilingual Capability by Cross-Lingual-Thought Prompting (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Large language models (LLMs) demonstrate impressive multilingual capability, but their performance varies substantially across different languages. |
| Approach: | They propose a generic template prompt that stimulates cross-lingual and logical reasoning skills to enhance task performance across languages. |
| Outcome: | The proposed method improves multilingual capability across languages and covers high-resource and low-resourced languages. |
Your Reasoning Benchmark May Not Test Reasoning: Revealing Perception Bottleneck in Abstract Reasoning Benchmarks (2026.acl-long)
Copied to clipboard
| Challenge: | Abstraction and Reasoning Corpus and ARC-AGI are widely used to assess progress in artificial intelligence. |
| Approach: | They propose a two-stage pipeline that separates perception and reasoning . they propose to test this pipeline against standard end-to-end one-stage evaluation . |
| Outcome: | The proposed pipeline separates perception and reasoning, and isolates reasoning from bottlenecks. |
Ling-CL: Understanding NLP Models through Linguistic Curricula (2023.emnlp-main)
Copied to clipboard
| Challenge: | linguistic complexity is a key component of learning in NLP, according to a new study . linguistic complex is based on lexical diversity, word sophistication, and readability . |
| Approach: | They employ a characterization of linguistic complexity from psycholinguistic and language acquisition research to develop data-driven curricula. |
| Outcome: | The proposed approach will inform future research in all NLP areas . it uses linguistic metrics (indices) that inform the challenges and reasoning required to address each task . |
Probabilistic Tree-of-thought Reasoning for Answering Knowledge-intensive Complex Questions (2023.findings-emnlp)
Copied to clipboard
| 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. |
Semantic-Aware Logical Reasoning via a Semiotic Framework (2026.acl-long)
Copied to clipboard
Yunyao Zhang, Xinglang Zhang, Junxi Sheng, Wenbing Li, Junqing Yu, Yi-Ping Phoebe Chen, Wei Yang, Zikai Song
| Challenge: | Existing studies largely overlook the interplay between logical complexity and semantic complexity, limiting their robustness under abstract propositions, ambiguous contexts, and conflicting stances. |
| Approach: | They propose a semiotic-square-guided framework that integrates automated deduction with reflective verification to manage logical complexity across deeper reasoning chains. |
| Outcome: | The proposed framework achieves state-of-the-art performance on RepublicQA with 6.25% average gain, and generalizes well to four mainstream logical reasoning benchmarks with an additional 7.05% improvement. |
EfficientXLang: Towards Improving Token Efficiency Through Cross-Lingual Reasoning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Despite recent advances in Reasoning Language Models, most research focuses solely on English, even though many models are pretrained on multilingual data. |
| Approach: | They evaluate three open-source RLMs: DeepSeek R1, Qwen 2.5, and Qwend 3 across four math datasets and seven typologically diverse languages. |
| Outcome: | The proposed model reduces token usage and preserves accuracy even after translation into English. |
DrAgent: Empowering Large Language Models as Medical Agents for Multi-hop Medical Reasoning (2025.findings-emnlp)
Copied to clipboard
Fenglin Liu, Zheng Li, Hongjian Zhou, Qingyu Yin, Jingfeng Yang, Xin Liu, Zhengyang Wang, Xianfeng Tang, Shiyang Li, Xiang He, Ruijie Wang, Bing Yin, Xiao Gu, Lei Clifton, David A. Clifton
| Challenge: | commercial LLMs can be difficult to use in real-world clinical decision-making . a lightweight LLM can be used to collaborate with diverse clinical tools . |
| Approach: | They propose a lightweight LLM that can be used to build medical LLMs as agents . they use recursive curriculum learning to optimize the LLM in an easy-to-hard progression . |
| Outcome: | The proposed approach outperforms human experts in medical examinations on diverse datasets. |
Debate4MATH: Multi-Agent Debate for Fine-Grained Reasoning in Math (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing data annotation methods suffer from high annotation cost and lack of effective automatic validation. |
| Approach: | They propose a Fine-grained Multi-Agent Debate framework and a dataset that prompts multiple agents to debate and then a Multi-agent Debates Reward Model (MRM) to improve its mathematical reasoning capabilities. |
| Outcome: | The proposed model outperforms the state-of-the-art methods by 1.2% and 3.5% on a GSM8K dataset and 45.1% on the MATH dataset. |
DetGPT: Detect What You Need via Reasoning (2023.emnlp-main)
Copied to clipboard
Renjie Pi, Jiahui Gao, Shizhe Diao, Rui Pan, Hanze Dong, Jipeng Zhang, Lewei Yao, Jianhua Han, Hang Xu, Lingpeng Kong, Tong Zhang
| Challenge: | Recent advances in the field of computer vision have enabled more effective and sophisticated interactions between humans and machines. |
| Approach: | They propose a reasoning-based object detection paradigm that leverages state-of-the-art multi-modal models and open-vocabulary object detectors to perform reasoning within the context of the user’s instructions and the visual scene. |
| Outcome: | The proposed method enables users to interact with the system using natural language instructions, allowing for a higher level of interactivity. |
Preemptive Answer “Attacks” on Chain-of-Thought Reasoning (2024.findings-acl)
Copied to clipboard
| 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. |
Causal-ESC: Reliable Policy Learning for Emotional Support Conversation via Causal Inference (2026.acl-long)
Copied to clipboard
| 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. |
Relation-Aware Question Answering for Heterogeneous Knowledge Graphs (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Existing retrieval-based approaches to solve multihop Knowledge Base Question Answering (KBQA) fail to utilize information from head-tail entities and the semantic connection between relations to enhance the information capturing of relations in KGs. |
| Approach: | They propose to use a dual relation graph to find the answer entity in a knowledge graph . they use primal entity graph reasoning, dual relation grafitment and interaction . |
| Outcome: | The proposed approach achieves significant performance gain over the prior state-of-the-art on two public datasets, WebQSP and CWQ. |
RARR: Researching and Revising What Language Models Say, Using Language Models (2023.acl-long)
Copied to clipboard
Luyu Gao, Zhuyun Dai, Panupong Pasupat, Anthony Chen, Arun Tejasvi Chaganty, Yicheng Fan, Vincent Zhao, Ni Lao, Hongrae Lee, Da-Cheng Juan, Kelvin Guu
| Challenge: | Language models (LMs) excel at many tasks but often produce unsupported or misleading content. |
| Approach: | They propose a system that finds attribution for any text generation model and post-edits it to fix unsupported content. |
| Outcome: | The proposed system improves attribution while preserving the original output. |
RSA-Bench: Benchmarking Audio Large Models in Real-World Acoustic Scenarios (2026.findings-acl)
Copied to clipboard
Yibo Zhang, Kaiwen Luo, Liang Lin, Shilinlu Yan, Jin Wang, Yaoqi Guo, Yitian Chen, Yalan Qin, Zhenhong Zhou, Kun Wang, Li Sun
| Challenge: | Existing evaluations rely on synthetic Gaussian noise or simplistic single-source interference, failing to capture the intricate, multi-layered acoustic dynamics that characterize authentic physical environments. |
| Approach: | They propose a robustness benchmark to stress-test Audio Large Models (ALLMs) using high-fidelity auditory scene simulations. |
| Outcome: | The proposed model performs well on a wide range of tasks, including automatic speech recognition, speech translation, and audio-based reasoning. |
Less is More: Improving LLM Reasoning with Minimal Test-Time Intervention (2026.acl-long)
Copied to clipboard
| Challenge: | Recent advances in large language models (LLMs) have focused on test-time scaling to improve reasoning quality but at the cost of efficiency. |
| Approach: | They propose a training-free framework that enhances reasoning accuracy and stability with minimal overhead. |
| Outcome: | The proposed framework yields consistent gains across general, coding, and STEM tasks while remaining highly efficient. |
Step-level Verifier-guided Hybrid Test-Time Scaling for Large Language Models (2025.emnlp-main)
Copied to clipboard
Kaiyan Chang, Yonghao Shi, Chenglong Wang, Hang Zhou, Chi Hu, Xiaoqian Liu, Yingfeng Luo, Yuan Ge, Tong Xiao, JingBo Zhu
| Challenge: | Recent training-based TTS methods, such as continued reinforcement learning, have surged in popularity, while training-free TTS approaches are gradually fading from prominence. |
| Approach: | They propose a fine-grained sequential scaling method guided by process verification that integrates training-free TTS methods with other classical parallel scaling methods at the step level. |
| Outcome: | Experiments on five instruction-tuned large language models (LLMs) show that training-free TTS methods can extend reasoning performance boundaries. |
CRAB: Assessing the Strength of Causal Relationships Between Real-world Events (2023.emnlp-main)
Copied to clipboard
| Challenge: | Existing models for reasoning about events in narratives do not understand the complexity of the causal relationships of events in the narrative. |
| Approach: | They propose a Causal Reasoning Assessment Benchmark to evaluate causal understanding of events in narratives. |
| Outcome: | The proposed model performs worse when models are derived from complex causal structures than simple linear causal chains. |
VideoCuRL: Video Curriculum Reinforcement Learning with Orthogonal Difficulty Decomposition (2026.acl-long)
Copied to clipboard
| Challenge: | Reinforcement Learning (RL) is crucial for Video-LLMs with complex spatiotemporal reasoning. |
| Approach: | They propose a framework that decomposes difficulty into two axes in video understanding . they employ efficient, training-free proxies to map data onto a 2D curriculum grid . |
| Outcome: | The proposed framework surpasses strong RL baselines on reasoning and perception tasks. |
Experience is the Teacher: Reusing Atomic Thoughts from LLMs to Improve Medical Dialogue (2026.findings-acl)
Copied to clipboard
Guangya Yu, Hui Luo, Qi Ye, Ruihui Hou, Weiyan Zhang, Mingxi Shang, Xuanwu Li, ChunMing Wang, Tong Ruan
| Challenge: | Recent large reasoning models (LLMs) lack dynamic and diverse thinking capabilities . reusing atomic thoughts provides a practical pathway toward dynamic reasoning . |
| Approach: | They propose a framework that extracts atomic thoughts from teacher models and reuses them to guide reasoning and generate responses. |
| Outcome: | The proposed framework extracts atomic thoughts from teacher models and reuses them to guide reasoning and generate responses. |
Quantifying and Mitigating Unimodal Biases in Multimodal Large Language Models: A Causal Perspective (2024.findings-emnlp)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models have facilitated the development of Multimodal LLMs. |
| Approach: | They propose a causal framework to interpret unimodal biases in visual question answering problems and a framework to integrate information from different modalities and mitigate biase. |
| Outcome: | The proposed framework analyzes visual question answering (VQA) problems to assess their impact on predictions. |
V-ALPHASOCIAL: Benchmark and Self-Reflective Chain-of-Thought Generation for Visual Social Commonsense Reasoning (2025.findings-acl)
Copied to clipboard
Zongyu Lin, Zhikun Xu, Xiaohan Song, Yixin Wan, Xingcheng Yao, Tsung-Han Lin, Selina Song, Pranav Subbaraman, Ben Zhou, Kai-Wei Chang, Yizhou Sun
| Challenge: | Social commonsense reasoning is a multimodal task that requires both textual and visual cues. |
| Approach: | They propose a method that integrates visual cues into social commonsense reasoning tasks. |
| Outcome: | The proposed method improves social commonsense reasoning on a multimodal foundation model. |
Tailoring Diagnostic Modeling to Individual Learners: Personalized Distractor Generation via MCTS-Guided Reasoning Reconstruction (2026.acl-long)
Copied to clipboard
| Challenge: | Current distractor generation methods produce shared distractors for all students, ignoring individual variations in reasoning, which limits their diagnostic effectiveness. |
| Approach: | They propose a method which tailors distractors to each student’s specific cognitive flaws, inferred from their past question-answering (QA) history. |
| Outcome: | The proposed framework outperforms existing methods in generating plausible distractors and adapts to group-level settings. |
Document-Level Relation Extraction with Global Relations and Entity Pair Reasoning (2025.findings-acl)
Copied to clipboard
| Challenge: | Existing document-level relation extraction models focus on individual entity pairs, limiting their ability to handle complex reasoning tasks. |
| Approach: | They propose a document-level relation extraction framework based on global relations and entity pair reasoning that captures fine-grained interactions between entity pairs. |
| Outcome: | The proposed framework outperforms existing models on widely-used datasets. |
Graph-GRPO: Stabilizing Multi-Agent Topology Learning via Group Relative Policy Optimization (2026.findings-acl)
Copied to clipboard
Yueyang Cang, Xiaoteng Zhang, Erlu Zhao, Zehua Ji, Yuhang Liu, Yuchen He, Zhiyuan Ning, Chen Yijun, Wenge Que, Li Shi
| Challenge: | Recent approaches to optimize communication topology rely on single-sample policy gradients with absolute rewards. |
| Approach: | They propose a topology optimization framework that integrates Group Relative Policy Optimization. |
| Outcome: | The proposed topology optimization framework outperforms state-of-the-art methods on reasoning and code generation benchmarks. |
Veri-R1: Toward Precise and Faithful Claim Verification via Online Reinforcement Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing approaches to online claim verification rely on prompt engineering or pre-designed reasoning workflows. |
| Approach: | They propose an online reinforcement learning framework that enables an LLM to interact with a search engine and receive reward signals that explicitly shape its planning, retrieval, and reasoning behaviors. |
| Outcome: | Empirical results show that Veri-R1 improves joint accuracy by 30% and doubles evidence score, often surpassing larger-scale model counterparts. |
Exploring the Numerical Reasoning Capabilities of Language Models: A Comprehensive Analysis on Tabular Data (2023.findings-emnlp)
Copied to clipboard
| Challenge: | Recent benchmarks have assessed language models' numerical abilities . limitations include tokenization and representation of numbers in text, hallucination, and a lack of numerical commonsense knowledge. |
| Approach: | They propose a hierarchical taxonomy for numerical reasoning skills that includes representation, number sense, manipulation, and complex reasoning. |
| Outcome: | The proposed model outperforms other models on the tabular Natural Language Inference task. |
PlanGEN: A Multi-Agent Framework for Generating Planning and Reasoning Trajectories for Complex Problem Solving (2025.emnlp-main)
Copied to clipboard
Mihir Parmar, Xin Liu, Palash Goyal, Yanfei Chen, Long Le, Swaroop Mishra, Hossein Mobahi, Jindong Gu, Zifeng Wang, Hootan Nakhost, Chitta Baral, Chen-Yu Lee, Tomas Pfister, Hamid Palangi
| Challenge: | Existing methods for natural planning lack constraint-guided iterative verification and adaptive selection . a recent study found that LLMs are not good at such planning. |
| Approach: | They propose a model-agnostic and easily scalable agent framework with three key components: constraint, verification, and selection agents. |
| Outcome: | The proposed framework improves inference-time algorithms on NATURAL PLAN and OlympiadBench benchmarks. |
CamoQuery: Language-Guided Reasoning Camouflaged Object Segmentation (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for camouflaged object segmentation are limited to vision-only mask prediction under fixed task assumptions. |
| Approach: | They propose a language-guided reasoning camouflaged object segmentation task that generates an intent-consistent segmentation mask from an image and an implicit query text instruction. |
| Outcome: | The proposed task can generate an intent-consistent segmentation mask from a camouflaged image and an implicit query text instruction. |
When Is Thinking Enough? Early Exit via Sufficiency Assessment for Efficient Reasoning (2026.acl-long)
Copied to clipboard
| 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. |
The Role of Visual Modality in Multimodal Mathematical Reasoning: Challenges and Insights (2025.acl-long)
Copied to clipboard
Yufang Liu, Yao Du, Tao Ji, Jianing Wang, Yang Liu, Yuanbin Wu, Aimin Zhou, Mengdi Zhang, Xunliang Cai
| Challenge: | Existing models that leverage visual information do not improve math reasoning performance . authors suggest that visual information is important for multimodal reasoning . |
| Approach: | They propose a dataset to require image reliance for problem-solving and challenge models with similar, yet distinct, images that change the correct answer. |
| Outcome: | The proposed model performance is unaffected by changes to or removal of images in the dataset. |
LLM-Guided Semantic Relational Reasoning for Multimodal Intent Recognition (2025.emnlp-main)
Copied to clipboard
| 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. |
Cognitive Scaffold: From Fluid Context to Crystallized Memory for Long-Horizon DeepResearch Agents (2026.acl-long)
Copied to clipboard
| Challenge: | Scaling LLM-based agents to long-horizon deep research is constrained by context-noise trade-off . solving a single query may require hundreds of interactions with noisy environments . |
| Approach: | They propose a factorized memory architecture that decouples the cognitive state into a Fluid Working Context for immediate reasoning and a persistent Knowledge Graph for long-term retention. |
| Outcome: | The Cognitive Scaffold outperforms baselines on Xbench-DeepSearch, BrowseComp-ZH, and GAIA . it achieves 74.7% Avg@3 and 87.0% Pass@3 on xbench, browseComp, and 88.3% Pass@3. |
S2G-RAG: Structured Sufficiency and Gap Judging for Iterative Retrieval-Augmented QA (2026.acl-long)
Copied to clipboard
| Challenge: | Retrieval-augmented generation grounds language models in external evidence, but multi-hop question answering remains difficult . iterative pipelines must control what to retrieve next and when evidence is adequate. |
| Approach: | They propose an iterative framework with an explicit controller, S2G-Judge . they map structured gap items into the next retrieval query to produce stable retrieval trajectories . |
| Outcome: | Experiments on TriviaQA, HotpotQA, and 2WikiMultiHopQA show that S2G-RAG improves multi-hop QA performance and robustness under multi-turn retrieval. |
Can Code-Switched Texts Activate a Knowledge Switch in LLMs? A Case Study on English-Korean Code-Switching (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Recent large language models (LLMs) demonstrate multilingual abilities, yet they are English-centric due to dominance of English in training corpora. |
| Approach: | They propose to use a synthetic English-korean CS question-answering dataset to investigate this potential. |
| Outcome: | The proposed model can activate, identify and leverage knowledge for reasoning in low-resource languages. |
a1: Steep Test-time Scaling Law via Environment Augmented Generation (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Language Models (LLMs) have made remarkable advances in reasoning, yet continue to struggle with hallucinations, logical errors, and inability to self-correct during complex multi-step tasks. |
| Approach: | They propose a framework that enhances LLM reasoning through real-time environmental feedback validating each reasoning step, dynamic branch exploration for investigating alternative solution paths when faced with errors, and experience-based learning from successful reasoning trajectories. |
| Outcome: | The proposed model outperforms comparable models by 24.4 percentage points across benchmarks while outperforming comparable models. |
MedAdapter: Efficient Test-Time Adaptation of Large Language Models Towards Medical Reasoning (2024.emnlp-main)
Copied to clipboard
| Challenge: | Large language models (LLMs) have improved generation and reasoning capabilities compared to traditional BERT-sized models due to massive number of parameters and extensive pre-training on vast textual corpora. |
| Approach: | They propose a unified post-hoc adapter for test-time adaptation of large language models . they propose to fine-tune only a small BERT-sized adapter to rank candidate LLMs . |
| Outcome: | The proposed adapter improves performance on four biomedical tasks without requiring computational resources or sharing data with third parties. |
Achieving Multi-Hop Calculation and Safe Abstention in Financial Numerical Reasoning by Metric Graph Constrained LLMs (2026.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) are prone to forced generation when confronting ambiguous evidence or complex recursive dependencies. |
| Approach: | They propose a framework that imposes semantic and structural constraints via a financial metric knowledge graph. |
| Outcome: | a neuro-symbolic framework outperforms existing models on financial metric knowledge graphs. |
Stepwise Informativeness Search for Improving LLM Reasoning (2025.emnlp-main)
Copied to clipboard
| Challenge: | Recent advances in large language models have improved multistep reasoning but they lose focus over the middle of long contexts. |
| Approach: | They propose a tree search framework that proactively identifies underutilized steps and minimizing redundant information between steps. |
| Outcome: | The proposed framework generates more accurate and concise rationales with reduced errors and redundancy. |
Do LLMs Overthink Basic Math Reasoning? Benchmarking the Accuracy-Efficiency Tradeoff in Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) achieve impressive performance on complex benchmarks yet sometimes fail on basic math reasoning. |
| Approach: | They propose a benchmark to evaluate the efficiency of reasoning in large language models . they formalize the accuracy-verbosity tradeoff and introduce the overthinking score . |
| Outcome: | The proposed model performs well on complex benchmarks but fails on basic math reasoning . the proposed model generates 18 more tokens while achieving lower accuracy . |
Empowering GraphRAG with Knowledge Filtering and Integration (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large language models suffer from knowledge gaps and hallucinations, resulting in incorrect or poor reasoning. |
| Approach: | They propose Graph retrieval-augmented generation (GraphRAG) which integrates structured knowledge from external graphs to enhance model's reasoning. |
| Outcome: | Experiments on knowledge graph QA tasks show that GraphRAG significantly improves reasoning performance across multiple backbone models. |
Mina: A Multilingual LLM-Powered Legal Assistant Agent for Empowering Access to Justice in Bangladesh (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing AI legal assistants lack Bengali-language support and jurisdiction-specific adaptation, limiting their effectiveness. |
| Approach: | They developed a multilingual LLM-based legal assistant tailored for the Bangladeshi context that employs multilingual embeddings and a RAG-based chain-of-tools framework for retrieval, reasoning, translation, and document generation. |
| Outcome: | Evaluated by law faculty from leading Bangladeshi universities across all stages of the 2022 and 2023 Bangladesh Bar Council examinations, Mina achieved scores of 75–80% in preliminary MCQs, written, and simulated viva voce components. |
Beyond Chunks and Graphs: Retrieval-Augmented Generation through Triplet-Driven Thinking (2026.findings-acl)
Copied to clipboard
| Challenge: | Retrieval-augmented generation (RAG) is critical for reducing hallucinations and incorporating external knowledge into Large Language Models (LLMs). |
| Approach: | They propose a framework that leverages an LLM to decompose questions into searchable triplets with placeholders. |
| Outcome: | Empirical results show that T2RAG outperforms state-of-the-art multi-round and Graph RAG methods while reducing retrieval costs by up to 45%. |
Exposing the Achilles’ Heel: Evaluating LLMs Ability to Handle Mistakes in Mathematical Reasoning (2025.acl-long)
Copied to clipboard
| Challenge: | Existing evaluations focus on final accuracy, neglecting the critical aspect of reasoning capabilities. |
| Approach: | They propose to evaluate LLMs’ abilities to detect and correct reasoning mistakes by using rule-based methods and smaller language models. |
| Outcome: | The proposed model outperforms existing models such as GPT-4o and GPT4 in both accuracy and accuracy, but lacks data contamination and memorization concerns. |
Under the Shadow of Babel: How Language Shapes Reasoning in LLMs (2025.findings-emnlp)
Copied to clipboard
| Challenge: | linguistic relativity suggests that the structure of language shapes cognitive patterns . large language models internalize the habitual logical structures embedded in different languages, authors say . |
| Approach: | a study introduces a bilingual dataset for causal reasoning in Chinese and English. |
| Outcome: | a new study shows that large language models internalize reasoning biases shaped by language . the model internalizes language-specific preferences and rigidly applies them to atypical inputs, the study shows . |
Stabilizing Efficient Reasoning with Step-Level Advantage Selection (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models generate long and verbose reasoning traces at inference time . short context post-training alone induces substantial reasoning compression . |
| Approach: | They propose a step-level advantage selection approach that reduces reasoning length by over 30% . they propose to use GRPO without any length-aware objective to train models in a shorter context window . |
| Outcome: | The proposed approach reduces average reasoning length by over 30% while improving Pass@1 accuracy by 3.79 points over the strongest length-aware baseline. |
CLARO: Controlled Attribute-Driven Reasoning Optimization for Efficient Chain-of-Thought (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent approaches guide reasoning length through token penalties or truncation, risking the omission of necessary steps. |
| Approach: | They propose a framework to internalize conciseness and attribute-driven reasoning optimization . they propose guiding models to embed high-quality structural attributes within a token budget . |
| Outcome: | The proposed method outperforms state-of-the-art models across benchmarks yielding accuracy gains of up to 63.6%. |
From Recall to Forgetting: Benchmarking Long-Term Memory for Personalized Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing benchmarks frame long-term memory evaluation as fact retrieval from past conversations, providing limited insight into agents’ ability to consolidate memory over time or handle frequent knowledge updates. |
| Approach: | They propose a long-term memory benchmark that evaluates three memory-grounded tasks: remembering, reasoning, and recommending. |
| Outcome: | The proposed benchmarks evaluate three tasks: remembering, reasoning, and recommending. |
Thinking Before You Speak: A Proactive Test-time Scaling Approach (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Large Language Models often exhibit deficiencies with complex reasoning tasks, such as maths, due to the discrepancy between human reasoning patterns and those presented in training data. |
| Approach: | They propose to insert insights between consecutive reasoning steps to bridge this gap by generating insights between the next reasoning steps. |
| Outcome: | Experiments on mathematical datasets confirm the effectiveness of the proposed reasoning framework on complex problems. |
Soft Head Selection for Injecting ICL-Derived Task Embeddings (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) are commonly adapted to downstream tasks using parameter-efficient fine-tuning (PEFT) or in-context learning (ICL). |
| Approach: | They propose a gradient-based method that derives task-specific embeddings from activations using few-shot prompts and injects them during inference. |
| Outcome: | The proposed method outperforms existing methods on open-ended generation, reasoning, and natural language understanding tasks while using fewer trainable parameters. |
ARM2: Adaptive Reasoning Model with Vision Understanding and Executable Code (2026.findings-acl)
Copied to clipboard
| Challenge: | Large Reasoning Models suffer from the "over-thinking" problem, causing performance degradation. |
| Approach: | They propose a unified model that balances reasoning performance and efficiency across multiple formats through a reinforcement learning framework augmented with length-aware optimization. |
| Outcome: | The proposed model reduces token costs while preserving performance compared to traditional models. |
Learning from Mistakes: Negative Reasoning Samples Enhance Out-of-Domain Generalization (2026.acl-long)
Copied to clipboard
Tian Xueyun, MingHua Ma, Bingbing Xu, Nuoyan Lyu, Wei Li, Heng Dong, Zheng Chu, Yuanzhuo Wang, Huawei Shen
| 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. |
Reasoning-Guided Exploration for Online DPO (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent work has aimed to enhance reasoning capabilities of language models, but these methods are limited to domains with objectively verifiable answers. |
| Approach: | They propose a self-play framework to improve reasoning on general-domain data. |
| Outcome: | Experiments show that the proposed framework improves reasoning performance on general-domain data while maintaining competitive performance on verifiable academic benchmarks. |
CRAFTQA: A Code-Driven Adaptive Framework for Complex Structured Data Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing unified structured data question answering methods rely on a set of predefined functions, which restricts their ability to perform complex reasoning beyond these predefined operations. |
| Approach: | They propose a novel adaptive code-driven framework that generates code-based reasoning operations based on a question. |
| Outcome: | The proposed framework improves on multiple structured datasets on real-world scenarios. |
Memorizing is Not Enough: Deep Knowledge Injection Through Reasoning (2025.acl-long)
Copied to clipboard
Ruoxi Xu, Yunjie Ji, Boxi Cao, Yaojie Lu, Hongyu Lin, Xianpei Han, Ben He, Yingfei Sun, Xiangang Li, Le Sun
| Challenge: | Existing knowledge injection frameworks focus on knowledge memorization and retrieval, but static nature of large language models leads to outdated information as the real world evolves or when adapting to domain-specific knowledge. |
| Approach: | They propose a four-tier knowledge injection framework that defines the levels of knowledge injection: memorization, retrieval, reasoning, and association. |
| Outcome: | The proposed framework defines the levels of knowledge injection: memorization, retrieval, reasoning, and association. |
Performance Gap in Entity Knowledge Extraction Across Modalities in Vision Language Models (2025.acl-long)
Copied to clipboard
| Challenge: | Vision-language models excel at extracting and reasoning about information from images, yet their capacity to leverage internal knowledge about specific entities remains underexplored. |
| Approach: | They propose a dataset which allows separating entity recognition and question answering . they hypothesize that this decline arises from limitations in how information flows from image tokens to query tokens. |
| Outcome: | The proposed model performance drops when the entity is presented visually rather than textually. |
HEALing Entropy Collapse: Enhancing Exploration in Few-Shot RLVR via Hybrid-Domain Entropy Dynamics Alignment (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for training reasoning-oriented large language models assume high-resource settings with abundant data. |
| Approach: | They propose a framework that integrates high-value general-domain data to promote more diverse exploration. |
| Outcome: | The proposed framework matches or surpasses RLVR trained with 32 target-domain samples using 32 target domain samples. |
ERRV: Eliciting Efficient Reasoning through Reasoning Vectors for Policy Optimization in Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing efforts to improve reasoning efficiency of large language models focus on modifying the reinforcement learning reward, such as adding length penalties. |
| Approach: | They propose a training framework that elicits efficient reasoning through reasoning vectors and a framework that allows the model to generate high-quality responses during reinforcement learning. |
| Outcome: | The proposed framework reduces reasoning length by 30% while maintaining stability, while retaining high accuracy. |
CARO: Chain-of-Analogy Reasoning Optimization for Robust Content Moderation (2026.findings-acl)
Copied to clipboard
| Challenge: | Current large language models struggle with ambiguous content moderation cases due to misleading "decision shortcuts" . authors propose a two-stage training framework to induce robust analogical reasoning in LLMs . |
| Approach: | They propose a two-stage training framework to induce robust analogical reasoning in LLMs . they bootstrap analogy reasoning chains via retrieval-augmented generation and SFT . |
| Outcome: | The proposed framework outperforms state-of-the-art reasoning models and specialized moderation models on ambiguous moderation benchmarks. |
QG-CoC: Question-Guided Chain-of-Captions for Large Multimodal Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing prompting methods for multimodal large language models lack fine-grained perception across disparate images . existing methods fail to integrate perception and reasoning, causing problems with general multi-image reasoning tasks. |
| Approach: | They propose a generalized prompting method that integrates perception and reasoning . they evaluate the method on open-source and closed-source MLLMs . |
| Outcome: | The proposed method shows competitive performance across tasks and improves in challenging scenarios. |
MM-ShiftKV: Decode-Aware Prefill-Stage KV Selection for Multimodal Large Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent work suggests a prefill-stage KV cache selection method to estimate KV importance from prefilling statistics. |
| Approach: | They propose a training-free, decode-aware and strictly prefill-only KV selection method that retains key-value caching for decoding . |
| Outcome: | The proposed method outperforms existing methods under tight cache budgets on multimodal benchmarks. |
Less is More: Knowledge-Aware Compression for Long Legal Judgment Prediction (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances leverage large language models (LLMs) for legal reasoning, but they face high computational costs and information degradation when handling long cases. |
| Approach: | They propose a framework that selectively retains legally relevant information while reducing redundant or less informative content, enabling efficient and accurate long-context reasoning. |
| Outcome: | The proposed framework outperforms existing methods on four real-world datasets spanning multiple jurisdictions and languages. |
Why Do LLM-based Web Agents Fail? A Hierarchical Planning Perspective (2026.acl-long)
Copied to clipboard
| Challenge: | Existing evaluations focus primarily on end-to-end success, offering limited insight into where failures arise. |
| Approach: | They propose a hierarchical planning framework that analyzes web agents across three layers . they show that structured Planning Domain Definition Language (PDDL) plans produce more concise and goal-directed strategies than natural language (NL) plans . |
| Outcome: | The proposed framework analyzes web agents across three layers to improve reasoning, grounding, and recovery. |
Reasoning-to-Defend: Safety-Aware Reasoning Can Defend Large Language Models from Jailbreaking (2025.emnlp-main)
Copied to clipboard
| Challenge: | Large Reasoning Models (LLMs) have demonstrated impressive performances across diverse domains, but how their safety benefits from enhanced reasoning capabilities against jailbreak queries remains unexplored. |
| Approach: | They propose a safety-aware reasoning paradigm that integrates a pivot token-based safety-based reasoning mechanism into LLMs’ generation process. |
| Outcome: | The proposed model improves the safety of large language models against jailbreak queries while minimizing attacks and maintaining the original performance. |
RouteRAG: Efficient Retrieval-Augmented Generation from Text and Graph via Reinforcement Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing graph-based or hybrid systems lack the ability to integrate supplementary evidence as reasoning unfolds. |
| Approach: | They propose a framework that integrates non-parametric knowledge into Large Language Models . they use a RL-based framework to optimize the entire generation process via RL . |
| Outcome: | The proposed framework outperforms existing RAG frameworks in five question answering benchmarks. |
Quantifying and Understanding Uncertainty in Large Reasoning Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for estimating generation uncertainty do not provide finite-sample guarantees for reasoning-answer generation. |
| Approach: | They propose a method that provides the uncertainty of the reasoning-answer structure with statistical guarantees. |
| Outcome: | The proposed method disentangles reasoning quality from answer correctness while establishing theoretical guarantees for efficient explanation methods. |
Reasoning in a Combinatorial and Constrained World: Benchmarking LLMs on Natural-Language Combinatorial Optimization (2026.findings-acl)
Copied to clipboard
| Challenge: | NLCO evaluates large language models for combinatorial optimization (CO) . existing evaluations emphasize relatively simple reasoning competencies . |
| Approach: | They propose a combinatorial optimization benchmark that evaluates large language models on CO reasoning. |
| Outcome: | The proposed model can handle combinatorial optimization without writing code or calling external solvers. |
DARL: Encouraging Diverse Answers for General Reasoning without Verifiers (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent efforts such as RLPR have extended RLVR to general domains, enabling training on broader datasets and achieving improvements over RL PR. |
| Approach: | They propose a framework that encourages the generation of diverse answers within a controlled deviation range from the reference while preserving alignment with it. |
| Outcome: | Extensive experiments on 13 benchmarks show that DARL surpasses RLPR in both reasoning accuracy and output diversity. |
PRISM: Probing Reasoning, Instruction, and Source Memory in LLM Hallucinations (2026.acl-long)
Copied to clipboard
Yuhe Wu, Guangyu Wang, Yuran Chen, Jiatong Zhang, Yutong Zhang, Yujie Chen, Jiaming Shang, Guang Zhang, Zhuang Liu
| Challenge: | Existing benchmarks for hallucination evaluation rely on mixed queries and posterior evaluation, which quantifies hallucinosity severity but offers limited insight into where and why they occur. |
| Approach: | They propose a controlled benchmark that disentangles hallucinations into four dimensions: knowledge missing, knowledge errors, reasoning errors, and instruction-following errors. |
| Outcome: | The proposed model disentangles hallucinations into four dimensions: knowledge missing, knowledge errors, reasoning errors, and instruction-following errors. |
Analyze Like a Venture Capitalist: Information-Gain and Knowledge Enhanced Graph Reasoning for Startup Success Prediction (2026.findings-acl)
Copied to clipboard
| Challenge: | Most venture capital investments fail, while a few deliver outsized returns. |
| Approach: | They propose a framework that synthesizes relational evidence across sources . they propose combining information-gain-driven retriever and knowledge base to ground reasoning . |
| Outcome: | The proposed framework achieves +5.9% F1 and +22.1% Precision@5 over state-of-the-art baselines. |
Think Smart, Not Hard: Difficulty Adaptive Reasoning for Large Audio Language Models (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing methods to determine whether to perform reasoning lack fine-grained mechanisms to adapt reasoning length to problem complexity. |
| Approach: | They propose a difficulty-adaptive reasoning method that dynamically links reasoning length to the model’s perceived problem difficulty. |
| Outcome: | The proposed method reduces average reasoning length by 50%, achieving higher efficiency without sacrificing accuracy. |
ThinkSLM: Towards Reasoning in Small Language Models (2025.emnlp-main)
Copied to clipboard
| Challenge: | Reasoning has long been viewed as an emergent property of large language models (LLMs), appearing at or above a certain scale (100B parameters). |
| Approach: | They propose a benchmark to evaluate the reasoning abilities of small language models (SLMs) using quantization, pruning, and distillation. |
| Outcome: | The proposed benchmark systematically evaluates reasoning abilities of small language models trained from scratch or derived from LLMs using quantization, pruning, and distillation. |
MATA: Multi-Agent Framework for Reliable and Flexible Table Question Answering (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in Large Language Models have significantly improved table understanding tasks . practical deployment of TableQA systems presents several persistent challenges . |
| Approach: | They propose a multi-agent TableQA framework that leverages multiple reasoning paths and tools built with small language models. |
| Outcome: | The proposed framework achieves state-of-the-art accuracy and efficient reasoning while avoiding excessive LLM inference. |
PICTURE: Enhancing Theory-of-Mind in Large Language Models by Revealing, Not Hiding, Characters’ Lack of Knowledge (2026.acl-long)
Copied to clipboard
| 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. |
Beyond Reasoning Gains: Mitigating General-Capability Forgetting in Large Reasoning Models (2026.findings-acl)
Copied to clipboard
Hoang Phan, Xianjun Yang, Yuanshun Yao, Jingyu Zhang, Shengjie Bi, Xiaocheng Tang, Madian Khabsa, Lijuan Liu, Deren Lei
| Challenge: | Reinforcement learning with verifiable rewards (RLVR) has delivered impressive gains in mathematical and multimodal reasoning . however, the recipe introduces a significant risk of capability regression, where models forget foundational skills after prolonged training without employing regularization strategies. |
| Approach: | They propose a replay strategy with dynamic objective reweighting for general knowledge preservation using short-horizon signals of convergence and instability. |
| Outcome: | The proposed method preserves general capabilities and improves reasoning . it can be applied to existing RLVR pipelines without training additional models or tuning . |
STARQA: A Question Answering Dataset for Complex Analytical Reasoning over Structured Databases (2025.emnlp-main)
Copied to clipboard
| Challenge: | Existing Text2SQL systems do not support SQL analysts in their primary work of performing complex analytics on specialized databases. |
| Approach: | They propose to decompose STARQA questions using SQL and Python to perform reasoning on specialized relational databases. |
| Outcome: | The proposed approach decomposes the task through a combination of SQL and Python, and achieves better performance on the more difficult questions. |
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
| 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. |
Adversarial Yet Cooperative: Multi-Perspective Reasoning in Retrieved-Augmented Language Models (2026.findings-acl)
Copied to clipboard
Can Xu, Lingyong Yan, Jiayi Wu, Haosen Wang, Shuaiqiang Wang, Yuchen Li, Jizhou Huang, Dawei Yin, Xiang Li
| Challenge: | Existing training paradigms rely on outcome-oriented rewards, which provide insufficient signal for shaping the complex, multi-step reasoning process. |
| Approach: | They propose a framework that integrates large reasoning models with retrieval-augmented generation to improve reasoning fidelity and verification rigor. |
| Outcome: | Experiments on multiple benchmarks demonstrate the effectiveness of the proposed framework. |
M3-VQA: A Benchmark for Multimodal, Multi-Entity, Multi-Hop Visual Question Answering (2026.acl-long)
Copied to clipboard
| Challenge: | Existing knowledge-based VQA benchmarks focus on coarse-grained categories and simple reasoning over single entities. |
| Approach: | They propose a knowledge-based Visual Question Answering benchmark to enhance multimodality evaluation. |
| Outcome: | The proposed benchmark improves evaluation of multimodal large language models in fine-grained multimodal entity understanding and complex multihop reasoning. |
CNSL-bench: Benchmarking the Sign Language Understanding Capabilities of MLLMs on Chinese National Sign Language (2026.acl-long)
Copied to clipboard
| Challenge: | CNSL-bench is the first comprehensive Chinese National Sign Language benchmark . current MLLMs are inferior to human performance, despite advances in multimodal modeling . |
| Approach: | They propose a Chinese National Sign Language benchmark to evaluate multimodal large language models in sign language understanding. |
| Outcome: | The proposed benchmark evaluates 21 open-source and proprietary MLLMs . results show that current models are inferior to human performance . |
KnowDR-REC: Auditing Knowledge-Conditioned Visual Grounding in Referring Expression Comprehension (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing evaluation metrics suggest that Multimodal large language models have acquired fine-grained visual grounding capabilities. |
| Approach: | They propose a benchmark to assess Referring Expression Comprehension (REC) that uses intra-image visual cues to localize target objects and a controllable evaluation mechanism to test sensitivity to fine-grained factual changes. |
| Outcome: | The proposed benchmarks show that multimodal large language models have a high level of performance on the RefCOCO family of benchmarks. |
Verified Critical Step Optimization for LLM Agents (2026.findings-acl)
Copied to clipboard
| Challenge: | Critical Step Optimization (CSO) focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success. |
| Approach: | They propose a method which focuses preference learning on verified critical steps where alternative actions demonstrably flip task outcomes from failure to success. |
| Outcome: | The proposed method outperforms the existing methods on GAIA-Text-103 and XBench-DeepSearch while requiring supervision at only 16% of trajectory steps. |
ReActR: Reasoning through Error-Activated Reflection for LLM Post-Training (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for improving the mathematical abilities of Large Language Models (LLMs) focus disproportionately on scaling correct training samples, overlooking the rich learning signals contained in erroneous reasoning trajectories. |
| Approach: | They propose a framework that enhances reasoning by learning reflective behaviors from erroneous trajectories by using data construction and training. |
| Outcome: | Extensive experiments on three LLMs show that ReActR improves reasoning performance on Llama-3-8B. |
LLMs as Lab Engineers: A Benchmark for Analytical Method Lifecycle Management (2026.findings-acl)
Copied to clipboard
Xiaoyi Chen, Mahsa Monshizadeh, Chaoqi Zhang, Jianjun Lang, Yang Wu, Genevieve Mortensen, Xiaozhong Liu, Haixu Tang
| Challenge: | General-purpose commercial models outperform domain-specialized ones, while RAG and reasoning significantly improve performance. |
| Approach: | They propose a benchmark to evaluate LLMs' capabilities in analytical chemistry scenarios. |
| Outcome: | The proposed framework outperforms existing benchmarks focused on factual knowledge and provides practical guidance for analytical chemistry challenges. |
AIPO: Adaptive Information Guided Token-Level Reinforcement Learning for Large Language Model Reasoning (2026.acl-long)
Copied to clipboard
| Challenge: | Existing RLVR methods focus on all generated tokens rather than on which tokens contribute to reasoning. |
| Approach: | They propose to use a Random–Fourier approximation of the Hilbert–Schmidt Independence Criterion to focus updates on decisive tokens discovered on the fly to improve the efficiency of mutual-information estimation. |
| Outcome: | The proposed approach yields +20% accuracy over strong RLVR baselines while updating merely 10% of tokens, demonstrating superior efficiency and effectiveness. |
AgentGym2: Benchmarking Large Language Model Agents in De-Idealized Real-World Environments (2026.acl-long)
Copied to clipboard
Zhiheng Xi, Dingwen Yang, Jiaqi Liu, Jixuan Huang, Honglin Guo, Baodai Huang, Tinggang Chen, Qi Zhang, Zhonghang Lu, Chenyu Liu, Jiajun Sun, Jiazheng Zhang, Dingwei Zhu, Xin Guo, Junzhe Wang, Zhihao Zhang, Yuming Yang, Junjie Ye, Minghe Gao, Dongrui Liu, Jiaming Ji, Guohao Li, Tao Gui, Qi Zhang, Xuanjing Huang
| Challenge: | Existing benchmarks evaluate agents in simplified, idealized settings, relying on pre-packaged tool interfaces, overlooking critical steps, and assume inputs are clean and fully specified. |
| Approach: | They propose a framework that evaluates language agents in simplified, idealized settings . they show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2 . |
| Outcome: | Experiments on 15 proprietary and open-source models show that even SOTA systems like Gemini and GPT-5 struggle on AgentGym2 . |
Beyond Fully Random Masking: Attention-Guided Denoising and Optimization for Diffusion Language Models (2026.acl-long)
Copied to clipboard
| Challenge: | Existing methods for full-attention dLLMs rely on random masking strategies that overlook intrinsic token dependencies. |
| Approach: | They propose an attention-guided denoising and optimization framework that aligns training and optimization with attention-derived dependencies. |
| Outcome: | The proposed framework outperforms state-of-the-art methods on mathematical and coding benchmarks. |
Can Small LLMs Learn a Robust Theory of Mind via RLVR? Investigating Generalization through the False-Belief Task (2026.findings-acl)
Copied to clipboard
| Challenge: | Recent advances in large language models (LLMs) have demonstrated emergent capabilities in complex reasoning, largely spurred by rule-based Reinforcement Learning (RL) techniques applied during post-training. |
| Approach: | They evaluate whether small-scale LLMs can acquire a robust and generalizable Theory of Mind (ToM) capability through RL with verifiable rewards. |
| Outcome: | The proposed model performs well on in-distribution tasks but fails to transfer to unseen ToM tasks with different characteristics. |
ReasonAny: Incorporating Reasoning Capability to Any Model via Simple and Effective Model Merging (2026.acl-long)
Copied to clipboard
| Challenge: | Existing models with long chain-of-thought reasoning lack reasoning depth and domain-specific utility. |
| Approach: | They propose a model merging framework that integrates reasoning with domain-specific task models. |
| Outcome: | The proposed model merging framework outperforms state-of-the-art models while maintaining robust reasoning performance. |
Reinforced Efficient Reasoning via Semantically Diverse Exploration (2026.acl-long)
Copied to clipboard
Ziqi Zhao, Zhaochun Ren, Jiahong Zou, Liu Yang, Zhiwei Xu, Xuri Ge, Zhumin Chen, Xinyu Ma, Daiting Shi, Shuaiqiang Wang, Dawei Yin, Xin Xin
| Challenge: | Existing methods for reinforcement learning with verifiable rewards suffer from limited exploration diversity and inefficient reasoning. |
| Approach: | They propose a method that rewards concise and correct reasoning while penalizing unnecessarily long reasoning chains. |
| Outcome: | Extensive experiments on Qwen and Llama models validate the effectiveness and efficiency of ROSE. |