Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhang, Kongcheng, Yao, Qi, Liu, Shunyu, Wang, Yingjie, Lai, Baisheng, Ye, Jieping, Song, Mingli, Tao, Dacheng
Natura: Preprint
Pubblicazione: 2025
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866910998857777152
author Zhang, Kongcheng
Yao, Qi
Liu, Shunyu
Wang, Yingjie
Lai, Baisheng
Ye, Jieping
Song, Mingli
Tao, Dacheng
author_facet Zhang, Kongcheng
Yao, Qi
Liu, Shunyu
Wang, Yingjie
Lai, Baisheng
Ye, Jieping
Song, Mingli
Tao, Dacheng
contents 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.
format Preprint
id arxiv_https___arxiv_org_abs_2506_08745
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning
Zhang, Kongcheng
Yao, Qi
Liu, Shunyu
Wang, Yingjie
Lai, Baisheng
Ye, Jieping
Song, Mingli
Tao, Dacheng
Artificial Intelligence
Computation and Language
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.
title Consistent Paths Lead to Truth: Self-Rewarding Reinforcement Learning for LLM Reasoning
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2506.08745