Large Language Models Have Intrinsic Meta-Cognition, but Need a Good Lens

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Ma, Ziyang, Yuan, Qingyue, Wang, Zhenglin, Zhou, Deyu
Format: Preprint
Published: 2025
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866918178036121600
author Ma, Ziyang
Yuan, Qingyue
Wang, Zhenglin
Zhou, Deyu
author_facet Ma, Ziyang
Yuan, Qingyue
Wang, Zhenglin
Zhou, Deyu
contents Previous research has primarily focused on the cognitive error detection capabilities of Large Language Models (LLMs), often prompting them to analyze mistakes in reasoning chains. However, few studies have examined the meta-cognitive abilities of LLMs (e.g., their self-awareness of step errors), which are crucial for their reliability. While studies on LLM self-evaluation present some measures, such as perplexity, which can reflect the answer correctness and be viewed as the lens of meta-cognition, they lack step-level analysis and adaptation. This paper studies the evaluation of LLM meta-cognition using the current lenses and how to improve these lenses. Specifically, we propose AutoMeco, an Automated Meta-cognition Evaluation framework for benchmarking the existing lenses. Furthermore, a training-free Markovian Intrinsic Reward Adjustment strategy, MIRA, is proposed to boost current meta-cognition lenses. Experimental results on three mathematical reasoning datasets and three LLMs show the reasonableness of AutoMeco by comparing it with Best-of-N verification. Moreover, the meta-cognition ability of LLMs can be better evaluated using MIRA.
format Preprint
id arxiv_https___arxiv_org_abs_2506_08410
institution arXiv
publishDate 2025
record_format arxiv
spellingShingle Large Language Models Have Intrinsic Meta-Cognition, but Need a Good Lens
Ma, Ziyang
Yuan, Qingyue
Wang, Zhenglin
Zhou, Deyu
Computation and Language
Previous research has primarily focused on the cognitive error detection capabilities of Large Language Models (LLMs), often prompting them to analyze mistakes in reasoning chains. However, few studies have examined the meta-cognitive abilities of LLMs (e.g., their self-awareness of step errors), which are crucial for their reliability. While studies on LLM self-evaluation present some measures, such as perplexity, which can reflect the answer correctness and be viewed as the lens of meta-cognition, they lack step-level analysis and adaptation. This paper studies the evaluation of LLM meta-cognition using the current lenses and how to improve these lenses. Specifically, we propose AutoMeco, an Automated Meta-cognition Evaluation framework for benchmarking the existing lenses. Furthermore, a training-free Markovian Intrinsic Reward Adjustment strategy, MIRA, is proposed to boost current meta-cognition lenses. Experimental results on three mathematical reasoning datasets and three LLMs show the reasonableness of AutoMeco by comparing it with Best-of-N verification. Moreover, the meta-cognition ability of LLMs can be better evaluated using MIRA.
title Large Language Models Have Intrinsic Meta-Cognition, but Need a Good Lens
topic Computation and Language
url https://arxiv.org/abs/2506.08410