Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning
This addresses the challenge of scalable reasoning training for LLMs, offering a method to reduce reliance on external supervision, though it is incremental as it builds on existing RL and consistency-based approaches.
The paper tackles the problem of training large language models for complex reasoning tasks without external supervision by proposing a self-rewarding reinforcement learning framework that uses consistency and volatility of intermediate reasoning states as intrinsic rewards, achieving performance comparable to or surpassing supervised RL on diverse benchmarks.
Recent advances of Reinforcement Learning (RL) have highlighted its potential in complex reasoning tasks, yet effective training often relies on external supervision, which limits the broader applicability. In this work, we propose a novel self-rewarding reinforcement learning framework to enhance Large Language Model (LLM) reasoning by leveraging the consistency of intermediate reasoning states across different reasoning trajectories. Our key insight is that correct responses often exhibit consistent trajectory patterns in terms of model likelihood: their intermediate reasoning states tend to converge toward their own final answers (high consistency) with minimal deviation toward other candidates (low volatility). Inspired by this observation, we introduce CoVo, an intrinsic reward mechanism that integrates Consistency and Volatility via a robust vector-space aggregation strategy, complemented by a curiosity bonus to promote diverse exploration. CoVo enables LLMs to perform RL in a self-rewarding manner, offering a scalable pathway for learning to reason without external supervision. Extensive experiments on diverse reasoning benchmarks show that CoVo achieves performance comparable to or even surpassing supervised RL. Our code is available at https://github.com/sastpg/CoVo.