Selective Expert Guidance for Effective and Diverse Exploration in Reinforcement Learning of LLMs

Fuente: arXiv
Guardado en:
Detalles Bibliográficos
Autores principales: Jiang, Zishang, Han, Jinyi, Li, Tingyun, Wang, Xinyi, Jiang, Sihang, Liang, Jiaqing, Dai, Zhaoqian, Ma, Shuguang, Yu, Fei, Xiao, Yanghua
Formato: Preprint
Publicado: 2025
Materias:
Acceso en línea:
Etiquetas: Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
_version_ 1866916990790139904
author Jiang, Zishang
Han, Jinyi
Li, Tingyun
Wang, Xinyi
Jiang, Sihang
Liang, Jiaqing
Dai, Zhaoqian
Ma, Shuguang
Yu, Fei
Xiao, Yanghua
author_facet Jiang, Zishang
Han, Jinyi
Li, Tingyun
Wang, Xinyi
Jiang, Sihang
Liang, Jiaqing
Dai, Zhaoqian
Ma, Shuguang
Yu, Fei
Xiao, Yanghua
contents Reinforcement Learning with Verifiable Rewards (RLVR) has become a widely adopted technique for enhancing the reasoning ability of Large Language Models (LLMs). However, the effectiveness of RLVR strongly depends on the capability of base models. This issue arises because it requires the model to have sufficient capability to perform high-quality exploration, which involves both effectiveness and diversity. Unfortunately, existing methods address this issue by imitating expert trajectories, which improve effectiveness but neglect diversity. To address this, we argue that the expert only needs to provide guidance only at critical decision points rather than the entire reasoning path. Based on this insight, we propose MENTOR: Mixed-policy Expert Navigation for Token-level Optimization of Reasoning, a framework that provides expert guidance only at critical decision points to perform effective and diverse exploration in RLVR. Extensive experiments show that MENTOR enables models capture the essence of expert strategies rather than surface imitation, thereby performing high-quality exploration and achieving superior overall performance. Our code is available online.
format Preprint
id arxiv_https___arxiv_org_abs_2510_04140
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Selective Expert Guidance for Effective and Diverse Exploration in Reinforcement Learning of LLMs
Jiang, Zishang
Han, Jinyi
Li, Tingyun
Wang, Xinyi
Jiang, Sihang
Liang, Jiaqing
Dai, Zhaoqian
Ma, Shuguang
Yu, Fei
Xiao, Yanghua
Artificial Intelligence
Computation and Language
Reinforcement Learning with Verifiable Rewards (RLVR) has become a widely adopted technique for enhancing the reasoning ability of Large Language Models (LLMs). However, the effectiveness of RLVR strongly depends on the capability of base models. This issue arises because it requires the model to have sufficient capability to perform high-quality exploration, which involves both effectiveness and diversity. Unfortunately, existing methods address this issue by imitating expert trajectories, which improve effectiveness but neglect diversity. To address this, we argue that the expert only needs to provide guidance only at critical decision points rather than the entire reasoning path. Based on this insight, we propose MENTOR: Mixed-policy Expert Navigation for Token-level Optimization of Reasoning, a framework that provides expert guidance only at critical decision points to perform effective and diverse exploration in RLVR. Extensive experiments show that MENTOR enables models capture the essence of expert strategies rather than surface imitation, thereby performing high-quality exploration and achieving superior overall performance. Our code is available online.
title Selective Expert Guidance for Effective and Diverse Exploration in Reinforcement Learning of LLMs
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2510.04140