paper-with-me

홈 › Papers

MoL-RL: Distilling Multi-Step Environmental Feedback into LLMs for Feedback-Independent Reasoning

2025-07-27 · Kang Yang, Jingxue Chen, Qingkun Tang, Tianxiang Zhang, Qianchun Lu arxiv

Large language models (LLMs) face significant challenges in effectively leveraging sequential environmental feedback (EF) signals, such as natural language evaluations, for feedback-independent chain-of-thought (CoT) reasoning. Existing approaches either convert EF into scalar rewards, losing rich contextual information, or employ refinement datasets, failing to exploit the multi-step and discrete nature of EF interactions. To address these limitations, we propose MoL-RL, a novel training paradigm that integrates multi-step EF signals into LLMs through a dual-objective optimization framework. Our method combines MoL (Mixture-of-Losses) continual training, which decouples domain-specific EF signals (optimized via cross-entropy loss) and general language capabilities (preserved via Kullback-Leibler divergence), with GRPO-based post-training to distill sequential EF interactions into single-step inferences. This synergy enables robust feedback-independent reasoning without relying on external feedback loops. Experimental results on mathematical reasoning (MATH-500, AIME24/AIME25) and code generation (CodeAgent-Test) benchmarks demonstrate that MoL-RL achieves state-of-the-art performance with the Qwen3-8B model, while maintaining strong generalization across model scales (Qwen3-4B). This work provides a promising approach for leveraging multi-step textual feedback to enhance LLMs' reasoning capabilities in diverse domains.

📄 PDF Abstract BibTeX arXiv:2507.20278

Code (0)

등록된 구현이 없습니다.

Tasks

Mathematical ReasoningCode Generation

Similar Papers 제목 키워드 기반

Distilling LLM Reasoning into an Interpretable Policy Tree for Human-AI Collaboration

2026-06-07 · Beiwen Zhang, Yongheng Liang, Guowei Zou, Haitao Wang 외 arxiv

Constructing efficient and reliable policies to assist humans is indispensable for human-AI collaboration. Existing methods mainly follow two lines of work. Most prior work relies on multi-agent reinforcement learning (M…

Multi-agent Reinforcement Learning

Experiential Reinforcement Learning

2026-02-15 · Taiwei Shi, Sihao Chen, Bowen Jiang, Linxin Song 외 arxiv

Reinforcement learning has become the central approach for language models (LMs) to learn from environmental reward or feedback. In practice, the environmental feedback is usually sparse and delayed. Learning from such s…

Reinforcement Learning

QueryAgent: A Reliable and Efficient Reasoning Framework with Environmental Feedback-based Self-Correction

2024-03-18 · Xiang Huang, Sitao Cheng, Shanshan Huang, Jiayu Shen 외

Employing Large Language Models (LLMs) for semantic parsing has achieved remarkable success. However, we find existing methods fall short in terms of reliability and efficiency when hallucinations are encountered. In thi…

Semantic Parsing

Learning with Rare Success but Rich Feedback via Reflection-Enhanced Self-Distillation

2026-05-12 · Yuwei Zhang, Sha Li, Changlong Yu, Qin Lu 외 arxiv

Enabling Large Language Models (LLMs) to continuously improve from environmental interactions is a central challenge in post-training. While on-policy self-distillation offers a promising paradigm, existing methods predo…

Continual Learning

Distilling Feedback into Memory-as-a-Tool

2026-01-09 · Víctor Gallego arxiv

We propose a framework that amortizes the cost of inference-time reasoning by converting transient critiques into retrievable guidelines, through a file-based memory system and agent-controlled tool calls. We evaluate th…