ReFLAIR: Enhancing Multimodal Reasoning via Structured Reflection and Reward-Guided Learning (2025.findings-emnlp)
Copied to clipboard
| Challenge: | Existing training methods for large models do not address the trade-off between reflection and accuracy. |
| Approach: | a unified framework teaches large models to perform structured reflection via an explicit $think re-think answer $ format and hybrid reward learning. |
| Outcome: | The proposed framework improves model performance on mathematical benchmarks and reduces inference cost by nearly 23%. |
Similar Papers
Vision-Language Models Can Self-Improve Reasoning via Reflection (2025.naacl-long)
Copied to clipboard
| Challenge: | Chain-of-thought (CoT) has been shown to improve the reasoning capability of large language models (LLMs). |
| Approach: | They propose a framework which iteratively enhances the model’s Vision-language Reasoning by Reflecting on CoT Rationales. |
| Outcome: | The proposed framework improves multimodal reasoning on vision-language tasks by 23% to 60% over baselines. |
Correct, Concise and Complete: Multi-stage Training For Adaptive Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Large language models (LLMs) increase test-time computation, often in the form of chain-of-thought (CoT) however, reasoning traces can become unnecessarily long, increasing computation costs without improving accuracy and sometimes even degrading performance. |
| Approach: | They propose a multi-stage efficient reasoning method that combines supervised fine-tuning with reinforcement learning using an adaptive length penalty. |
| Outcome: | The proposed method reduces response length by an average of 28% for 8B models and 40% for 32B models while incurring only minor performance drops of 1.6 and 2.5 points, respectively. |
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. |
CLARity: Reasoning Consistency Alone Can Teach Reinforced Experts (2026.acl-long)
Copied to clipboard
| Challenge: | Existing solutions to supervise the reasoning process are prohibitively expensive. |
| Approach: | They propose a cost-effective reinforcement learning framework that enhances reasoning quality using a small, general-purpose LLM only. |
| Outcome: | Experiments show that CLARity improves reasoning quality by 16.5% over standard outcome-based reinforcement learning (RL) human evaluations confirm substantial gains in factual correctness and reasoning coherence, leading to more trustworthy model outputs. |
AutoRubric: Rubric-Based Generative Rewards for Faithful Multimodal Reasoning (2026.findings-acl)
Copied to clipboard
| Challenge: | Multimodal large language models (MLLMs) have advanced from perception tasks to complex multi-step reasoning. |
| Approach: | They propose a framework that integrates reinforcement learning with verifiable rewards with process-level supervision through automatically collected rubric-based generative rewards. |
| Outcome: | The proposed framework achieves state-of-the-art performance on six multimodal reasoning benchmarks and significantly improves reasoning faithfulness in dedicated evaluations. |
ReflexiCoder: Teaching Large Language Models to Self-Reflect on Generated Code and Self-Correct It via Reinforcement Learning (2026.findings-acl)
Copied to clipboard
| Challenge: | Existing iterative refinement strategies that generate solutions in a single forward pass often hit a performance ceiling on complex algorithmic tasks. |
| Approach: | They propose a reinforcement learning framework that internalizes the structured reasoning trajectory directly into the model’s weights. |
| Outcome: | The proposed framework achieves 94.51% (87.20%) on HumanEval, 81.80% (78.57%) on MBPP, 35.00% on BigCodeBench, 52.21% on LiveCodeBech, and 37.34% on CodeForces in a single-attempt setting. |
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. |
Learning from Cognition: Enhancing RL Efficiency for LLM Reasoning via Hierarchical Metacognitive Decomposition and Refinement (2026.acl-long)
Copied to clipboard
Zexu Sun, Yongcheng Zeng, Erxue Min, Heyang Gao, Bokai Ji, Dugang Liu, Xing Tang, Xiuqiang He, Xu Chen
| Challenge: | Recent advances in Large Language Models have demonstrated notable inferential capacities via reinforcement learning (RL) however, “zero-RL” approaches relying on fixed prompt templates introduce substantial sampling inefficiencies for weak LLMs. |
| Approach: | They propose a hierarchical metacognitive RL framework that decomposes zero-accuracy problems into subproblems and prompts the policy to refine answers by referencing previous wrong solutions. |
| Outcome: | The proposed framework improves sample utilization and sample efficiency and accelerates convergence compared to baselines. |
Mirror: Multiple-perspective Self-Reflection Method for Knowledge-rich Reasoning (2024.acl-long)
Copied to clipboard
| Challenge: | Large language models (LLMs) struggle with knowledge-rich problems without external resources. |
| Approach: | They propose a Multiple-perspective self-reflection method that allows LLMs to reflect from multiple-perceptive clues, achieved through a heuristic interaction between a Navigator and a Reasoner. |
| Outcome: | The proposed method is superior to other self-reflection methods on five reasoning datasets. |
ReflectEvo: Improving Meta Introspection of Small LLMs by Learning Self-Reflection (2025.findings-acl)
Copied to clipboard
Jiaqi Li, Xinyi Dong, Yang Liu, Zhizhuo Yang, Quansen Wang, Xiaobo Wang, Song-Chun Zhu, Zixia Jia, Zilong Zheng
| Challenge: | ReflectEvo-460k is a large-scale, comprehensive, self-generated reflection dataset with broadened instructions and diverse multi-domain tasks. |
| Approach: | They propose a pipeline that iteratively generates self-reflection for self-training and a large-scale reflection dataset with broadened instructions and diverse multi-domain tasks. |
| Outcome: | The proposed pipeline improves Llama-3 reasoning ability by up to 71.2% and Mistral by upto 44.4%. |