REINFORCEMENT LEARNING (RL

DeepResearcher: Scaling Deep Research via Reinforcement Learning in Real-world Environments (2025.emnlp-main)

Challenge:	e-commerce payment fraud detection is a new area for reinforcement learning (RL) and Large Language Models (LLMs).
Approach:	They propose to integrate reinforcement learning (RL) with Large Language Models (LLMs) by framing transaction risk as a multi-step Markov Decision Process (MDP), RL optimizes risk detection across multiple payment stages.
Outcome:	The proposed approach improves fraud detection accuracy and demonstrates zero-shot capability.

Copied to clipboard

Yuxiang Zheng, Dayuan Fu, Xiangkun Hu, Xiaojie Cai, Lyumanshan Ye, Pengrui Lu, Pengfei Liu

Learning to Ideate for Machine Learning Engineering Agents (2026.eacl-short)

Challenge:	Large Language Models (LLMs) with web search capabilities show significant potential for deep research.
Approach:	They introduce a framework for end-to-end training of LLM-based deep research agents . they implement a specialized multi-agent architecture where browsing agents extract relevant information from various webpage structures.
Outcome:	The proposed framework improves on open-domain research tasks by 28.9 points over prompt engineering and 7.2 points over RAG-based RL agents.

Copied to clipboard

Yunxiang Zhang, Kang Zhou, Zhichao Xu, Kiran Ramnath, Yun Zhou, Sangmin Woo, Haibo Ding, Lin Lee Cheong

Challenge:	Existing machine learning engineering (MLE) agents struggle to iteratively optimize their implemented algorithms for effectiveness.
Approach:	They propose a framework that separates ideation from implementation that allows an implementation agent to request strategic help from a dedicated Ideator.
Outcome:	The proposed framework outperforms implementation-only agent baselines on MLE-Bench and can be trained with reinforcement learning to generate more effective ideas.

Reinforced Cross-modal Alignment for Radiology Report Generation (2022.findings-acl)

Copied to clipboard

Han Qin, Yan Song

Context Reasoner: Incentivizing Reasoning Capability for Contextualized Privacy and Safety Compliance via Reinforcement Learning (2025.emnlp-main)

Challenge:	Medical images are widely used in clinical decision-making, where writing radiology reports can be enhanced by automatic solutions to alleviate physicians’ workload.
Approach:	They propose an approach with reinforcement learning over a cross-modal memory to better align visual and textual features for radiology report generation.
Outcome:	The proposed approach improves cross-modal alignment on two English radiology report datasets and human evaluation confirms the results.

Copied to clipboard

Wenbin Hu, Haoran Li, Huihao Jing, Qi Hu, Ziqian Zeng, Sirui Han, Xu Heli, Tianshu Chu, Peizhao Hu, Yangqiu Song

Self-imitation Learning for Action Generation in Text-based Games (2023.eacl-main)

Challenge:	Current mitigation strategies fail to preserve contextual reasoning capabilities in risky scenarios, leading to systemic risks for legal compliance.
Approach:	They propose to use reinforcement learning with a rule-based reward to incentivize contextual reasoning capabilities while enhancing compliance with safety and privacy norms.
Outcome:	The proposed model outperforms Qwen2.5-7B-Instruct model in safety and privacy benchmarks and achieves +8.58% accuracy improvement.

Copied to clipboard

Zijing Shi, Yunqiu Xu, Meng Fang, Ling Chen

Using Semantic Similarity as Reward for Reinforcement Learning in Sentence Generation (P19-2)

Challenge:	Text-based games are situated systems where the game agents observe textual descriptions, and generate textual commands to interact with the environment.
Approach:	They propose a confidence-based self-imitation model to generate action candidates for the RL agent by exploiting past valuable trajectories to adapt a pre-trained language model towards a target game.
Outcome:	The proposed model performs well in multiple challenging games.

Copied to clipboard

Go Yasui, Yoshimasa Tsuruoka, Masaaki Nagata

Semantic Text Processing

Challenge:	Existing models for sentence generation use cross-entropy loss as the loss function . however, cross-etropy is unable to evaluate sentences as a whole and lacks flexibility . et al., 2018: a novel approach to improve sentence generation models .
Approach:	They propose a method to train a model using estimated semantic similarity between output and reference sentences to alleviate cross-entropy loss problems.
Outcome:	The proposed model improves the BLEU scores from the baseline LSTM NMT model.

Prompt-Based Length Controlled Generation with Multiple Control Types (2024.findings-acl)

Copied to clipboard

Renlong Jie, Xiaojun Meng, Lifeng Shang, Xin Jiang, Qun Liu

Low-Resource NLP

Challenge:	Existing length control methods focus on a simple control type of “equal to” a target length.
Approach:	They propose a prompt-based method to achieve length controlled generation under different control types with high accuracy by using reinforcement learning and sample filtering with the reward signal given by rule-based reward models.
Outcome:	The proposed method significantly improves the accuracy of prompt-based length control on popular summarization datasets like CNNDM and NYT under multiple control types.

Datasets and Recipes for Video Temporal Grounding via Reinforcement Learning (2025.emnlp-industry)

Copied to clipboard

Ruizhe Chen, Tianze Luo, Zhiting Fan, Heqing Zou, Zhaopeng Feng, Guiyang Xie, Hansheng Zhang, Zhuochen Wang, Zuozhu Liu, Zhang Huaijian

Visual Data in NLP

Challenge:	Existing methods for video temporal grounding suffer from limited temporal awareness and poor generalization.
Approach:	They propose a two-stage training framework that integrates supervised fine-tuning with reinforcement learning to improve both the accuracy and robustness of VTG models.
Outcome:	The proposed training framework outperforms existing models on multiple benchmarks on open-domain and challenging scenarios.

J4R: Learning to Judge with Equivalent Initial State Group Relative Policy Optimization (2026.acl-long)

Copied to clipboard

Austin Xu, Yilun Zhou, Xuan-Phi Nguyen, Caiming Xiong, Shafiq Joty

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

Challenge:	Large language models (LLMs) are increasingly being used for reasoning intensive tasks.
Approach:	They propose an algorithm that trains judges to be robust to positional biases . they also propose a benchmark that evaluates judges in diverse reasoning settings .
Outcome:	The proposed algorithm outperforms GPT-4o and the next best small judge by 6.7% and 9% on ReasoningJudgeBench and JudgeBench.

Copied to clipboard

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

ChatR1: Reinforcement Learning for Conversational Reasoning and Retrieval Augmented Question Answering (2026.acl-long)

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

Copied to clipboard

Simon Lupart, Mohammad Aliannejadi, Evangelos Kanoulas

Natural Language Interfaces

Challenge:	Unlike static ‘rewrite, retrieve, and generate’ pipelines, ChatR1 interleaves search and reasoning across turns, enabling exploratory and adaptive behaviors learned through RL.
Approach:	They propose a reasoning framework based on reinforcement learning (RL) for conversational question answering that interleaves search and reasoning across turns and provides turn-level feedback.
Outcome:	The proposed framework outperforms competing models on five CQA datasets, measured by different metrics (F1, BERTScore, and LLM-as-judge).

Mixture-of-Minds: Multi-Agent Reinforcement Learning for Table Understanding (2026.acl-long)

Copied to clipboard

Yuhang Zhou, Mingrui Zhang, Ke Li, Mingyi Wang, Qiao Liu, Qifei Wang, Jiayi Liu, Fei Liu, Serena Li, Weiwei LI, Mingze Gao, Abhishek Kumar, Xiangjun Fan, Zhuokai Zhao, Lizhu Zhang

Eliciting Medical Reasoning with Knowledge-enhanced Data Synthesis: A Semi-Supervised Reinforcement Learning Approach (2026.findings-acl)

Challenge:	Large language models (LLMs) have shown promise on understanding and reasoning over tables, but current approaches remain limited.
Approach:	They propose a multi-agent framework that decomposes table reasoning into three specialized roles: planning, coding, and answering.
Outcome:	The proposed framework decomposes table reasoning into three specialized roles: planning, coding, and answering.

Copied to clipboard

Haolin Li, Shuyang Jiang, Ruipeng Zhang, Jiangchao Yao, Ya Zhang, Yanfeng Wang

Semantic Text Processing

Challenge:	Existing methods to enhance medical reasoning lack high-quality data.
Approach:	They propose a medical knowledge-enhanced data Synthesis and Semi-supervised Reinforcement learning framework that uses rare disease knowledge to synthesize distribution-controllable reasoning questions.
Outcome:	The proposed method outperforms existing methods across ten medical benchmarks and achieves up to 5.93% gain on rare diseases tasks.

Speaking at the Right Level: Literacy-Controlled Counterspeech Generation with RAG-RL (2025.findings-emnlp)

Copied to clipboard

Xiaoying Song, Anirban Saha Anik, Dibakar Barua, Pengcheng Luo, Junhua Ding, Lingzi Hong

To Code or not to Code? Adaptive Tool Integration for Math Language Models via Expectation-Maximization (2025.findings-acl)

Challenge:	Existing approaches produce uniform responses, ignoring that health literacy levels affect the accessibility and effectiveness of counterspeech.
Approach:	They propose a Controlled-Literacy framework that generates counterspeech adapted to different health literacy levels.
Outcome:	The proposed framework outperforms baselines by generating more accessible counterspeech to health misinformation.

Copied to clipboard

Haozhe Wang, Long Li, Chao Qu, Weidi Xu, Fengming Zhu, Wei Chu, Fangzhen Lin

SPaCe: Unlocking Sample-Efficient Large Language Models Training With Self-Pace Curriculum Learning (2026.findings-acl)

Challenge:	Existing tools that integrate chain-of-thought reasoning and code execution lack metacognitive awareness to integrate tools.
Approach:	They propose a framework that synergizes structured exploration with off-policy RL optimization to create a cycle between metacognitive tool-use decisions and evolving capabilities.
Outcome:	The proposed framework improves over 11% on MATH500 and 9.4% on AIME without o1-like CoT.

Copied to clipboard

Van Dai Do, Manh Nguyen, Svetha Venkatesh, Hung Le

LongReward: Improving Long-context Large Language Models with AI Feedback (2025.acl-long)

Challenge:	Existing training pipelines sample training examples uniformly across steps or epochs, ignoring differences in difficulty, redundancy, and learning value, which slows learning and wastes computation.
Approach:	They propose a self-paced learning framework that enables efficient learning based on the capability of the model being trained through optimizing which data to use and when.
Outcome:	The proposed framework achieves comparable or better accuracy than state-of-the-art baselines while using up to (100 times) fewer samples.

Copied to clipboard

Jiajie Zhang, Zhongni Hou, Xin Lv, Shulin Cao, Zhenyu Hou, Yilin Niu, Lei Hou, Yuxiao Dong, Ling Feng, Juanzi Li

Controlling Multimodal Conversational Agents with Coverage-Enhanced Latent Actions (2026.acl-long)

Challenge:	In recent years, significant advancements have been achieved in the development of long-context large language models (LLMs).
Approach:	They propose a method that utilizes an off-the-shelf LLM to provide rewards for long-context model responses from four human-valued dimensions: helpfulness, logicality, faithfulness, and completeness.
Outcome:	The proposed method improves models’ long-context performance and enhances their ability to follow short instructions.

Copied to clipboard

Yongqi Li, Hao Lang, Tieyun Qian, Yongbin Li

Semantic Text Processing

Challenge:	Recent reinforcement learning (RL) has been widely explored for adapting MCAs to various human-AI interaction scenarios.
Approach:	They propose to use a latent action space for reinforcement learning instead of RL to fine-tune MCAs.
Outcome:	The proposed method outperforms baselines on two conversation tasks with a novel cycle consistency loss.

Instructions are all you need: Self-supervised Reinforcement Learning for Instruction Following (2026.acl-long)

Copied to clipboard

Qingyu Ren, Qianyu He, Powei Chang, Jie Zeng, Zeye Sun, Fei Yu, Jiaqing Liang, Yanghua Xiao

AgentOCR: Reimagining Agent History via Optical Self-Compression (2026.acl-long)

Challenge:	Existing reinforcement learning approaches suffer from dependency on external supervision and sparse reward signals from multi-constraint tasks.
Approach:	They propose a self-supervised reinforcement learning framework that eliminates dependency on external supervision by deriving reward signals directly from instructions and generating pseudo-labels for reward model training.
Outcome:	The proposed framework achieves strong improvements across 3 in-domain and 5 out-of-domain datasets while maintaining computational efficiency.

Copied to clipboard

Lang Feng, Fuchao Yang, Feng Chen, Xin Cheng, Haiyang Xu, Zhenglin Wan, Ming Yan, Bo An

Visual Data in NLP

Challenge:	Recent advances in large language models (LLMs) have enabled agentic systems trained with reinforcement learning over multi-turn interaction, but practical deployment is bottlenecked by rapidly growing textual histories that inflate token and memory costs.
Approach:	They propose a framework that represents the accumulated observation-action history as a compact rendered image.
Outcome:	The proposed framework preserves over 95% of text-based agent performance while significantly reducing token consumption (>50%), yielding consistent token and memory efficiency.

Neuro-Symbolic Approaches for Text-Based Policy Learning (2021.emnlp-main)

Copied to clipboard

Subhajit Chaudhury, Prithviraj Sen, Masaki Ono, Daiki Kimura, Michiaki Tatsubori, Asim Munawar

Explainability & Interpretability in NLP

Challenge:	Text-based games are important testbeds for reinforcement learning in the natural language domain.
Approach:	They propose a method that learns interpretable action policy rules from symbolic abstractions of textual observations for improved generalization.
Outcome:	The proposed method outperforms existing methods in RL using 5-10x fewer training games.

SALT: Step-level Advantage Assignment for Long-horizon Agents via Trajectory Graph (2026.findings-eacl)

Copied to clipboard

Jiazheng Li, Yawei Wang, Qiaojing Yan, Yijun Tian, Zhichao Xu, Huan Song, Panpan Xu, Lin Lee Cheong

Structured Data in NLP

Challenge:	Large Language Models (LLMs) have demonstrated remarkable capabilities, but their application to complex, multi-step, and long-horizon tasks remains challenging.
Approach:	They propose a framework that provides a finer-grained advantage assignment derived solely from outcome rewards.
Outcome:	The proposed framework provides a finer-grained advantage assignment, derived solely from outcome rewards.

Prejudge-Before-Think: Enhancing Large Language Models at Test-Time by Process Prejudge Reasoning (2025.findings-emnlp)

Copied to clipboard

Jianing Wang, Jin Jiang, Yang Liu, Mengdi Zhang, Xunliang Cai

Refine and Imitate: Reducing Repetition and Inconsistency in Persuasion Dialogues via Reinforcement Learning and Human Demonstration (2021.findings-emnlp)

Challenge:	Experimental results from competition-level complex reasoning demonstrate that bootstrapping with process prejudge can significantly enhance the reasoning ability of LLMs.
Approach:	They propose a new process prejudge strategy for LLM reasoning that bootstraps with process prejudgment .
Outcome:	The proposed method can be bootstrapped with process prejudge in LLM reasoning . it allows the model to anticipate errors rather than relying on trial and error.

Copied to clipboard

Weiyan Shi, Yu Li, Saurav Sahay, Zhou Yu

Dialogue Systems & Conversational Agents

Challenge:	Persuasion dialogue systems have long-standing problems of dialogue repetition and inconsistency which could impact user experience and impede the persuaded outcome.
Approach:	They propose to refine a language model baseline without user simulators and distill sentence-level information about repetition, inconsistency, and task relevance through rewards.
Outcome:	The proposed model outperforms state-of-the-art models on automatic metrics and human evaluation results on a donation persuasion task and generates more diverse, consistent and persuasive conversations according to user feedback.

Source Critical Reinforcement Learning for Transferring Spoken Language Understanding to a New Language (C18-1)

Copied to clipboard

He Bai, Yu Zhou, Jiajun Zhang, Liang Zhao, Mei-Yuh Hwang, Chengqing Zong

Information Extraction & Text Mining

Challenge:	a study aims to develop a language transferring system to avoid the trouble of acquiring and labeling a new big SLU corpus . general-purpose translators cannot handle the lot of semantic labels, not to mention cultural differences . a RL-based language transfer method can be used to adapt the adapted translator to a target language .
Approach:	They propose to use reinforcement learning to adapt a spoken language understanding model to a target language.
Outcome:	The proposed language transferring method improves domain classification accuracy by 22% compared with naive translation . the proposed language transfer method can be used on Chinese to English translators with more proper slot tags .

Incentivizing Strong Reasoning from Weak Supervision (2026.eacl-long)

Copied to clipboard

Yige Yuan, Teng Xiao, Shuchang Tao, Xue Wang, Jinyang Gao, Bolin Ding, Bingbing Xu

Ignore the KL Penalty! Boosting Exploration on Critical Tokens to Enhance RL Fine-Tuning (2025.findings-naacl)

Challenge:	Large language models (LLMs) have demonstrated impressive performance on reasoning-intensive tasks, but enhancing their reasoning abilities typically relies on expensive high-quality demonstrations and reinforcement learning.
Approach:	They propose to incentivize reasoning abilities of large language models without expensive demonstrations and reinforcement learning.
Outcome:	The proposed model can recover 94% of the gains of expensive RL at a fraction of the cost.

Copied to clipboard

Jean Vassoyan, Nathanaël Beau, Roman Plaud

Boundary-Guided Policy Optimization for Memory-efficient RL of Diffusion Large Language Models (2026.acl-long)

Challenge:	Existing studies show that large language models can learn to achieve long-term goals when fine-tuned with reinforcement learning.
Approach:	They propose a modification to the Kullback-Leibler penalty to favor exploration on critical tokens . they show how varying degrees of pre-training influence exploration .
Outcome:	The proposed model favors exploration on critical tokens, increasing the efficiency of the RL fine-tuning stage.

Copied to clipboard

Nianyi Lin, Jiajie Zhang, Lei Hou, Juanzi Li

NeoAMT: Neologism-Aware Agentic Machine Translation with Reinforcement Learning (2026.acl-long)

Challenge:	Existing methods for reinforcement learning (RL) require a large sample size to be implemented.
Approach:	They propose a memory-efficient RL algorithm that maximizes a lower bound of the ELBO-based objective.
Outcome:	Experiments show that BGPO outperforms previous RL algorithms for diffusion large language models in math problem solving, code generation, and planning tasks.

Copied to clipboard

Zhongtao Miao, Kaiyan Zhao, Masaaki Nagata, Yoshimasa Tsuruoka

Machine Translation

Challenge:	Neologism-aware machine translation aims to translate source sentences containing neologismes into target languages.
Approach:	They propose an agentic framework for neologism-aware machine translation equipped with a Wiktionary-based search toolkit.
Outcome:	The proposed framework is based on a Wiktionary-based search toolkit and a dedicated dataset for neologism-aware machine translation.

NAT: Enhancing Agent Tuning with Negative Samples (2025.naacl-long)

Copied to clipboard

Renxi Wang, Xudong Han, Yixuan Zhang, Timothy Baldwin, Haonan Li

Challenge:	Existing methods for fine-tuning and reinforcement learning use only positive examples, limiting their efficiency in low-resource scenarios.
Approach:	They propose a method that leverages both successful and failed trajectories for fine-tuning, maximizing the utility of limited resources.
Outcome:	The proposed method surpasses existing methods, including SFT, DPO, and PPO, across various tasks.

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

Copied to clipboard

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

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

Deep Reinforcement Learning with Hierarchical Action Exploration for Dialogue Generation (2024.lrec-main)

Copied to clipboard

Itsugun Cho, Ryota Takahashi, Yusaku Yanase, Hiroaki Saito

Dialogue Systems & Conversational Agents

Challenge:	Existing approaches to improve dialogues with random sampling are inefficient due to the large number of eligible responses with high action values.
Approach:	They propose a dual-granularity Q-function that extracts actions based on a grained hierarchy . they use offline RL and learn from multiple reward functions designed to capture emotional nuances in human interactions.
Outcome:	The proposed approach outperforms baselines across automatic metrics and human evaluations.

Agentic Conversational Search with Contextualized Reasoning via Reinforcement Learning (2026.findings-acl)

Copied to clipboard

Fengran Mo, Yifan Gao, Sha Li, Hansi Zeng, Xin Liu, Zhaoxuan Tan, Xian Li, Jianshu Chen, Dakuo Wang, Meng Jiang

Natural Language Interfaces

Challenge:	Existing studies focus on single-turn scenarios, which might lack the ability to handle multi-turn interactions.
Approach:	They propose a conversational agent that interleaves search and reasoning across turns and provides tailored rewards towards evolving user goals.
Outcome:	The proposed agent interleaves search and reasoning across turns, enabling exploratory and adaptive behaviors learned through reinforcement learning (RL) training with tailored rewards towards evolving user goals.

AlphaQuanter: An End-to-End Tool-Augmented Agentic Reinforcement Learning Framework for Stock Trading (2026.findings-acl)

Copied to clipboard

Zheye Deng, Weixiang Yan, Changlong Yu, Jiashu Wang

Look Again, Think Slowly: Enhancing Visual Reflection in Vision-Language Models (2025.emnlp-main)

Challenge:	Existing Large Language Models (LLMs) lack the end-to-end optimization needed to learn a coherent strategy from market feedback.
Approach:	They propose a single-agent framework that uses reinforcement learning to learn a dynamic policy over a transparent decision workflow.
Outcome:	The proposed framework achieves state-of-the-art performance on key financial metrics.

Copied to clipboard

Pu Jian, Junhong Wu, Wei Sun, Chen Wang, Shuo Ren, Jiajun Zhang

Towards Pareto-Efficient RLHF: Paying Attention to a Few High-Reward Samples with Reward Dropout (2024.findings-emnlp)

Challenge:	Recent advances in text-only "slow thinking" reasoning have prompted efforts to transfer this capability to vision-language models (VLMs).
Approach:	They propose a VRM Reflection-V which enhances visual reflection based on reasoning data for cold-start and reward design for reinforcement learning.
Outcome:	The proposed model improves visual reflection for cold-start and reward design for reinforcement learning (RL) it maintains a stronger and more consistent reliance on visual information during visual reasoning, indicating effective enhancement in visual reflection capabilities.

Copied to clipboard

Changhun Lee, Chiehyeon Lim

Green & Sustainable NLP

Challenge:	RLHF is a bi-objective problem that has the nature of a Pareto optimization . reward dropout is generalizable and most effective with non-pretrained target models .
Approach:	They propose a method that guarantees a Pareto improvement by leveraging reinforcement learning to fine-tune language models.
Outcome:	The proposed method guarantees a Pareto improvement on two benchmark datasets . it is generalizable and most effective with non-pretrained target models, saving the effort of pretraining.

Selective Token Generation for Few-shot Natural Language Generation (2022.coling-1)

Copied to clipboard

Daejin Jo, Taehwan Kwon, Eun-Sol Kim, Sungwoong Kim

Demystifying Multilingual Reasoning in Process Reward Modeling (2025.findings-emnlp)

Challenge:	Experimental results show that the proposed selective token generation algorithm outperforms the previous additive learning algorithms based on the PLMs.
Approach:	They propose an additive learning algorithm that selectively outputs language tokens between a task-general PLM and a specific adapter during training and inference.
Outcome:	The proposed algorithm outperforms existing methods on few-shot natural language generation tasks.

Copied to clipboard

Weixuan Wang, Minghao Wu, Barry Haddow, Alexandra Birch

SafeMCP: Proactive Power Regulation for LLM Agent Defense via Environment-Grounded Look-Ahead Reasoning (2026.acl-long)

Challenge:	a recent study focuses on the use of large language models to solve multi-step reasoning tasks.
Approach:	They propose to extend large language models to multilingual settings by extending process reward models to English . they train multilingual PRMs on a dataset spanning seven languages, which is translated from english .
Outcome:	The proposed model improves accuracy and reduces early-stage reasoning errors.

Copied to clipboard

Lichao Wang, ZhaoXing Ren, Tianzhuo Yang, Jiaming Ji, Chi Harold Liu, Yaodong Yang, Juntao Dai

Reasoning for Hierarchical Text Classification: The Case of Patents (2026.findings-acl)

Challenge:	Large Language Model (LLM) agents are expanding their action spaces to operate in complex environments.
Approach:	They propose a server-side defense plugin that constrains tool acquisition via predictive reasoning regarding future safety risks.
Outcome:	Experiments on PowerSeeking Bench, ToolEmu, and AgentHarm show that SafeMCP achieves a safe equilibrium, effectively mitigating risks while preserving agent utility.

Copied to clipboard

Lekang Jiang, Wenjun Sun, Stefan Goetz

Information Extraction & Text Mining

Challenge:	Hierarchical text classification (HTC) is one of the hardest HTC scenarios because of professional difficulties and extensive labels.
Approach:	They propose a framework that reformulates hierarchical classification as a step-by-step reasoning task.
Outcome:	The proposed framework outperforms supervised fine-tuning benchmarks on other widely used HTC benchmarks.

DORA: A Dual-Objective Reinforcement Learning Framework for Effective and Efficient Multimodal Agentic Search (2026.acl-long)

Copied to clipboard

Guangming Qin, Yuhao Deng, Yukun Zhao, Zhenyang Li, Junfeng Wang, Dawei Yin, Ye Yuan, Guoren Wang, Yizhou Yan, Chengliang Chai, Lei Cao

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

Challenge:	Existing methods to train large language models overlook quality of intermediate search results . existing methods often invoke search calls during reasoning, making inference inefficient .
Approach:	They propose a dual-objective reinforcement learning framework to improve search strategies of MLLMs . DORA outperforms state-of-the-art methods, achieving up to 8.4% higher accuracy .
Outcome:	The proposed model outperforms state-of-the-art methods while reducing search calls by 9.7%.

Copied to clipboard

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

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

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

Copied to clipboard

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

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

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

Copied to clipboard

Yining Huang, Bin Li, Keke Tang, Meilian Chen

Countering Reward Over-Optimization in LLM with Demonstration-Guided Reinforcement Learning (2024.findings-acl)

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

Copied to clipboard

Mathieu Rita, Florian Strub, Rahma Chaabouni, Paul Michel, Emmanuel Dupoux, Olivier Pietquin

Challenge:	Existing approaches address ROO by adding KL regularization, requiring computationally expensive hyperparameter tuning.
Approach:	They propose a reinforcement learning approach that leverages human demonstrations and a reward model to recalibrate the reward objective.
Outcome:	The proposed approach achieves comparable performance to carefully tuned baselines while mitigating ROO in three RL language tasks.

Self-EmoQ: Plutchik-Guided Value-based Planning to Drive Streaming Emotional TTS (2026.findings-acl)

Copied to clipboard

Yue Zhao, Hongyan Li, Yong Chen, Luo Ji

Towards Medical Complex Reasoning with LLMs through Medical Verifiable Problems (2025.findings-acl)

Challenge:	Existing systems lack a self-emotion determination mechanism to drive the streaming text-to-speech (TTS) synthesis.
Approach:	They propose an emotion-planning framework that determines the emotion prior to the textual generation, grounding the downstream emotional TTS in a streaming manner.
Outcome:	The proposed framework outperforms baselines on DailyDialog, EmoryNLP, IMEOCAP, and MELD on emotional alignment, contextual coherence, and expressive fluency.

Copied to clipboard

Junying Chen, Zhenyang Cai, Ke Ji, Xidong Wang, Wanlong Liu, Rongsheng Wang, Benyou Wang

Feasible is Not Enough: Cost-Aware Optimal Tool-Chain Planning on Multi-Solution Tool Graphs (2026.findings-acl)

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.

Copied to clipboard

Liangliang Liu, Yanming Li, Yigang Liu, Jialong Han, Rujia Shen, Yi Guan, Yi Lin, Jingchi Jiang

Green & Sustainable NLP

Challenge:	Existing tools and benchmarks often form tool learning (TL) as a single-solution setting . exploring large-scale TG is computationally expensive, especially under constrained context budgets.
Approach:	They propose a framework for learning optimal TL policies over large tool graphs . they train a reinforcement learning agent to acquire transferable expansion skills .
Outcome:	The proposed framework improves task success and solution optimality by 46.21% and 66.34% on multiSoTLBench.

SGG-R 3: From Next-Token Prediction to End-to-End Unbiased Scene Graph Generation (2026.findings-acl)

Copied to clipboard

Jiaye Feng, Qixiang Yin, Yuankun Liu, Tong Mo, Weiping Li

DVCQR: Dual-View Conversational Query Rewriting with Stage-wise Reinforcement Learning (2026.acl-long)

Challenge:	Existing methods for scene graph generation lack task-specific structured reasoning and sparse, long-tailed relation distributions.
Approach:	They propose a structured reasoning framework that integrates task-specific Chain-of-Thought and reinforcement learning with group sequence policy optimization to achieve unbiased scene graph generation.
Outcome:	The proposed framework achieves superior performance on two benchmarks.

Copied to clipboard

Chenyi Li, Xinhui Tu, Zaixiang Wang

Natural Language Interfaces

Challenge:	Recent approaches to improve retrieval effectiveness rely on a single rewrite . however, they often suffer from conflicting optimization signals .
Approach:	They propose a dual-view CQR framework that generates two complementary rewrites for each query.
Outcome:	Experiments show that DVCQR outperforms state-of-the-art methods on most metrics . the proposed framework generates two complementary rewrites for each query .

SAIR-Comb : A Structure-Aware Iterative Refinement Framework for Combinatorics Autoformalization (2026.acl-long)

Copied to clipboard

Weijie Jiang, Gaolei He, Beibei Xiong, Jianlin Wang, Zhengfeng Yang

Syntactic Text Processing

Challenge:	Large language models (LLMs) have catalyzed advances in mathematical reasoning, propelling the development of automated theorem proving (ATP).
Approach:	They propose a Structure-Aware Iterative Refinement framework for Combinatorics powered by Lean 4 and LLMs.
Outcome:	The proposed framework achieves strong performance on the specialized CombiBench while remaining highly competitive on general-domain benchmarks.

CSPO: Alleviating Reward Ambiguity for Structured Table-to-LaTeX Generation (2026.acl-long)

Copied to clipboard

Yunfan Yang, Cuiling Lan, Jitao Sang, Yan Lu

Enhancing Study-Level Inference from Clinical Trial Papers via Reinforcement Learning-Based Numeric Reasoning (2025.emnlp-main)

Challenge:	Tables contain rich structured information, but when stored as images their contents remain "locked" within pixels.
Approach:	They propose a framework that disentangles optimization across LaTeX tables components . CSPO assigns component-specific rewards and backpropagates each signal through tokens .
Outcome:	The proposed framework disentangles optimization across LaTeX tables components—structure, style, and content.

Copied to clipboard

Massimiliano Pronesti, Michela Lorandi, Paul Flanagan, Oisín Redmond, Anya Belz, Yufang Hou

MUSEG: Reinforcing Video Temporal Understanding via Timestamp-Aware Multi-Segment Grounding (2026.acl-long)

Challenge:	Prior work has framed this task as a textual inference task by retrieving relevant content fragments and inferring conclusions from them.
Approach:	They propose to extract structured numerical evidence and apply domain knowledge informed logic to derive outcome-specific conclusions.
Outcome:	The proposed approach outperforms general-purpose LLMs of over 400B parameters and achieves a 21% improvement in F1 score over retrieval-based systems.

Copied to clipboard

Fuwen Luo, Shengfeng Lou, Chi Chen, Ziyue Wang, Chenliang Li, Weizhou Shen, Jiyue Guo, Peng Li, Ming Yan, Ji Zhang, Fei Huang, Yang Liu

Visual Data in NLP

Challenge:	Existing methods for MLLMs struggle with fine-grained temporal reasoning . despite advances in video understanding, current methods struggle with time-sensitive tasks .
Approach:	They propose a time-stamp-aware multi-segment grounding method that enhances temporal understanding by introducing timestamps.
Outcome:	The proposed method outperforms existing methods on time-sensitive tasks and generalizes well across diverse temporal understanding scenarios.

ShopSimulator: Evaluating and Exploring RL-Driven LLM Agent for Shopping Assistants (2026.acl-long)

Copied to clipboard

Pei Wang, Yanan Wu, Xiaoshuai Song, Weixun Wang, Gengru Chen, Zhongwen Li, Kezhong Yan, Qi Liu, Ken Deng, Shuaibing Zhao, Shaopan Xiong, Xuepeng Liu, Xuefeng Chen, Wanxi Deng, Wenbo Su, Bo Zheng

Challenge:	Existing studies on large language model-based agents focus on evaluation benchmarks without training support.
Approach:	They propose a large-scale Chinese shopping simulation environment that uses large language models to train agents.
Outcome:	The proposed model performs poorly in a large-scale and challenging shopping environment in China.

TexOCR: Advancing Document OCR Models for Compilable Page-to-LaTeX Reconstruction (2026.acl-long)

Copied to clipboard

Chengye Wang, Lin Fu, Zexi Kuang, Yilun Zhao

GSM-Noise: Exploring and Enhancing Large Language Models’ Reasoning under Noisy Inputs (2026.findings-acl)

Challenge:	Existing document OCR largely targets plain text or Markdown, discarding structural and executable properties that make LaTeX essential for scientific publishing.
Approach:	They propose a benchmark and a training corpus for document reconstruction . they train a 2B-parameter model using supervised fine-tuning and reinforcement learning .
Outcome:	The proposed model improves on existing models using supervised fine-tuning and reinforcement learning with verifiable rewards.

Copied to clipboard

Zhengxin Zhang, Chengyu Huang, Xufu Liu, Dan Zhao, Jinyan Su, Claire Cardie

Robustness in NLP

Challenge:	Large language models struggle when dealing with complex, ill-formed, or noisy inputs . open-source models are less robust, while closed-source ones are more robust .
Approach:	They propose to use GSM-Noise to refine inputs before engaging in in-depth analysis to improve LLM robustness under noisy conditions.
Outcome:	The proposed model can achieve consistent performance gains under noisy conditions with prompt engineering, supervised finetuning, and reinforcement learning.

Robust Tool Use via Fission-GRPO: Learning to Recover from Execution Errors (2026.acl-long)

Copied to clipboard

Zhiwei Zhang, Fei Zhao, Rui Wang, Zezhong Wang, Bin Liang, Jiakang Wang, Yao Hu, Shaosheng Cao, Kam-Fai Wong

Robustness in NLP

Challenge:	Large language models (LLMs) can call tools effectively, but they remain brittle in multi-turn execution.
Approach:	They propose a framework that converts execution errors into on-policy corrective supervision within the RL training loop.
Outcome:	The proposed framework improves the error recovery rate of Qwen3-8B by 5.7% absolute and overall accuracy by 4.0% on BFCL v4 Multi-Turn.

rSIM: Incentivizing Reasoning Capabilities of LLMs via Reinforced Strategy Injection (2026.acl-long)

Copied to clipboard

Sijia Chen, Baochun Li, Di Niu

WebAnchor: Anchoring Agent Planning to Stabilize Long-Horizon Web Reasoning (2026.findings-acl)

Challenge:	Existing literature on Reasoning Language Models (RLMs) focuses on the ability to integrate reasoning strategies into the chain-of-thought process, contributing to improved problem-solving accuracy.
Approach:	They propose a reinforced strategy injection mechanism that enables any LLM to become an RLM by employing a small planner to guide the LLM's CoT through the adaptive injection of reasoning strategies.
Outcome:	The proposed model outperforms existing models in mathematical, coding, and financial reasoning tasks and is generalizable.

Copied to clipboard

Xinmiao Yu, Liwen Zhang, Xiaocheng Feng, Pengjun Xie, Jingren Zhou, Bing Qin, Yong Jiang

Reinforcement Learning for Diffusion LLMs via Energy-Based Gibbs Alignment (2026.acl-long)

Challenge:	Existing methods for reinforcement learning (RL)-based agents struggle with long-horizon planning and strategy coherence.
Approach:	They propose a reinforcement learning framework that decouples planning and execution.
Outcome:	The proposed framework outperforms baseline and first-step RL frameworks on four benchmarks.

Copied to clipboard

Yijia Fan, Jing Yang, Mingyu Liu, Kaitong Cai, Jian Wang, Keze Wang, Jusheng Zhang