Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Yang, Junyao, Qian, Chen, Wang, Kun, Zhang, Linfeng, Zhang, Quanshi, Liu, Yong, Liu, Dongrui
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866916039495778304
author Yang, Junyao
Qian, Chen
Wang, Kun
Zhang, Linfeng
Zhang, Quanshi
Liu, Yong
Liu, Dongrui
author_facet Yang, Junyao
Qian, Chen
Wang, Kun
Zhang, Linfeng
Zhang, Quanshi
Liu, Yong
Liu, Dongrui
contents The advancement of Large Reasoning Models (LRMs) has catalyzed a paradigm shift from reactive ``fast thinking'' text generation to systematic, step-by-step ``slow thinking'' reasoning, unlocking state-of-the-art performance in complex mathematical and logical tasks. However, the field faces \textit{the fundamental gap between token-level behavioral analysis and internal reasoning mechanisms, and the instability of reinforcement learning (RL) for reasoning optimization relying on costly external verifiers}. We identify and formally define \textbf{Entropy-Gradient Inversion}, a robust negative correlation between token entropy and logit gradients that acts as a definitive geometric fingerprint for LRM reasoning capability. Building on this, we propose \textbf{Correlation-Regularized Group Policy Optimization (CorR-PO)}, which embeds this inversion signature into RL reward regularization. Extensive experiments on various reasoning benchmarks across multiple model scales show CorR-PO consistently outperforms state-of-the-art baselines, confirming that stronger inversion directly correlates with superior reasoning performance.
format Preprint
id arxiv_https___arxiv_org_abs_2605_17770
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models
Yang, Junyao
Qian, Chen
Wang, Kun
Zhang, Linfeng
Zhang, Quanshi
Liu, Yong
Liu, Dongrui
Artificial Intelligence
Computation and Language
The advancement of Large Reasoning Models (LRMs) has catalyzed a paradigm shift from reactive ``fast thinking'' text generation to systematic, step-by-step ``slow thinking'' reasoning, unlocking state-of-the-art performance in complex mathematical and logical tasks. However, the field faces \textit{the fundamental gap between token-level behavioral analysis and internal reasoning mechanisms, and the instability of reinforcement learning (RL) for reasoning optimization relying on costly external verifiers}. We identify and formally define \textbf{Entropy-Gradient Inversion}, a robust negative correlation between token entropy and logit gradients that acts as a definitive geometric fingerprint for LRM reasoning capability. Building on this, we propose \textbf{Correlation-Regularized Group Policy Optimization (CorR-PO)}, which embeds this inversion signature into RL reward regularization. Extensive experiments on various reasoning benchmarks across multiple model scales show CorR-PO consistently outperforms state-of-the-art baselines, confirming that stronger inversion directly correlates with superior reasoning performance.
title Entropy-Gradient Inversion: Moving Toward Internal Mechanism of Large Reasoning Models
topic Artificial Intelligence
Computation and Language
url https://arxiv.org/abs/2605.17770