Efficient Reinforcement Learning with Semantic and Token Entropy for LLM Reasoning

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Cao, Hongye, Bai, Zhixin, Peng, Ziyue, Wang, Boyan, Yang, Tianpei, Huo, Jing, Zhang, Yuyao, Gao, Yang
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866914257995563008
author Cao, Hongye
Bai, Zhixin
Peng, Ziyue
Wang, Boyan
Yang, Tianpei
Huo, Jing
Zhang, Yuyao
Gao, Yang
author_facet Cao, Hongye
Bai, Zhixin
Peng, Ziyue
Wang, Boyan
Yang, Tianpei
Huo, Jing
Zhang, Yuyao
Gao, Yang
contents Reinforcement learning with verifiable rewards (RLVR) has demonstrated superior performance in enhancing the reasoning capability of large language models (LLMs). However, this accuracy-oriented learning paradigm often suffers from entropy collapse, which reduces policy exploration and limits reasoning capabilities. To address this challenge, we propose an efficient reinforcement learning framework that leverages entropy signals at both the semantic and token levels to improve reasoning. From the data perspective, we introduce semantic entropy-guided curriculum learning, organizing training data from low to high semantic entropy to guide progressive optimization from easier to more challenging tasks. For the algorithmic design, we adopt non-uniform token treatment by imposing KL regularization on low-entropy tokens that critically impact policy exploration and applying stronger constraints on high-covariance portions within these tokens. By jointly optimizing data organization and algorithmic design, our method effectively mitigates entropy collapse and enhances LLM reasoning. Experimental results across 6 benchmarks with 3 different parameter-scale base models demonstrate that our method outperforms other entropy-based approaches in improving reasoning.
format Preprint
id arxiv_https___arxiv_org_abs_2512_04359
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Efficient Reinforcement Learning with Semantic and Token Entropy for LLM Reasoning
Cao, Hongye
Bai, Zhixin
Peng, Ziyue
Wang, Boyan
Yang, Tianpei
Huo, Jing
Zhang, Yuyao
Gao, Yang
Artificial Intelligence
Reinforcement learning with verifiable rewards (RLVR) has demonstrated superior performance in enhancing the reasoning capability of large language models (LLMs). However, this accuracy-oriented learning paradigm often suffers from entropy collapse, which reduces policy exploration and limits reasoning capabilities. To address this challenge, we propose an efficient reinforcement learning framework that leverages entropy signals at both the semantic and token levels to improve reasoning. From the data perspective, we introduce semantic entropy-guided curriculum learning, organizing training data from low to high semantic entropy to guide progressive optimization from easier to more challenging tasks. For the algorithmic design, we adopt non-uniform token treatment by imposing KL regularization on low-entropy tokens that critically impact policy exploration and applying stronger constraints on high-covariance portions within these tokens. By jointly optimizing data organization and algorithmic design, our method effectively mitigates entropy collapse and enhances LLM reasoning. Experimental results across 6 benchmarks with 3 different parameter-scale base models demonstrate that our method outperforms other entropy-based approaches in improving reasoning.
title Efficient Reinforcement Learning with Semantic and Token Entropy for LLM Reasoning
topic Artificial Intelligence
url https://arxiv.org/abs/2512.04359