arXiv:2602.02605 · v3

Fine-Tuning Language Models to Know What They Know

首次提交 2026-02-02 12:08

版本更新 2026-11-11 04:13

雷达观测 2026-11-13 11:12

摘要

Evaluating true metacognition in Large Language Models (LLMs) is difficult due to biases and heuristics. This paper presents a framework to measure and enhance LLM metacognition while controlling for these biases. A measurement method using the $d'_{\rm type2}$ metric is established to isolate metacognitive ability. The Evolution Strategy for Metacognitive Alignment (ESMA) is proposed, demonstrating robust generalization across unseen datasets, languages, and newly acquired knowledge. Gradual parameter updates reveal that these improvements are driven by a sparse set of parameters, opening a potential avenue for targeted metacognitive optimization.

作者备注：Accepted to NeurIPS 2026