Large Language Model Agents Are Not Always Faithful Self-Evolvers

Fuente: arXiv
Salvato in:
Dettagli Bibliografici
Autori principali: Zhao, Weixiang, Wang, Yingshuo, Zhang, Yichen, Deng, Yang, Zhao, Yanyan, Che, Wanxiang, Qin, Bing, Liu, Ting
Natura: Preprint
Pubblicazione: 2026
Soggetti:
Accesso online:
Tags: Aggiungi Tag
Nessun Tag, puoi essere il primo ad aggiungerne!!
_version_ 1866908818963693568
author Zhao, Weixiang
Wang, Yingshuo
Zhang, Yichen
Deng, Yang
Zhao, Yanyan
Che, Wanxiang
Qin, Bing
Liu, Ting
author_facet Zhao, Weixiang
Wang, Yingshuo
Zhang, Yichen
Deng, Yang
Zhao, Yanyan
Che, Wanxiang
Qin, Bing
Liu, Ting
contents Self-evolving large language model (LLM) agents continually improve by accumulating and reusing past experience, yet it remains unclear whether they faithfully rely on that experience to guide their behavior. We present the first systematic investigation of experience faithfulness, the causal dependence of an agent's decisions on the experience it is given, in self-evolving LLM agents. Using controlled causal interventions on both raw and condensed forms of experience, we comprehensively evaluate four representative frameworks across 10 LLM backbones and 9 environments. Our analysis uncovers a striking asymmetry: while agents consistently depend on raw experience, they often disregard or misinterpret condensed experience, even when it is the only experience provided. This gap persists across single- and multi-agent configurations and across backbone scales. We trace its underlying causes to three factors: the semantic limitations of condensed content, internal processing biases that suppress experience, and task regimes where pretrained priors already suffice. These findings challenge prevailing assumptions about self-evolving methods and underscore the need for more faithful and reliable approaches to experience integration.
format Preprint
id arxiv_https___arxiv_org_abs_2601_22436
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Large Language Model Agents Are Not Always Faithful Self-Evolvers
Zhao, Weixiang
Wang, Yingshuo
Zhang, Yichen
Deng, Yang
Zhao, Yanyan
Che, Wanxiang
Qin, Bing
Liu, Ting
Computation and Language
Self-evolving large language model (LLM) agents continually improve by accumulating and reusing past experience, yet it remains unclear whether they faithfully rely on that experience to guide their behavior. We present the first systematic investigation of experience faithfulness, the causal dependence of an agent's decisions on the experience it is given, in self-evolving LLM agents. Using controlled causal interventions on both raw and condensed forms of experience, we comprehensively evaluate four representative frameworks across 10 LLM backbones and 9 environments. Our analysis uncovers a striking asymmetry: while agents consistently depend on raw experience, they often disregard or misinterpret condensed experience, even when it is the only experience provided. This gap persists across single- and multi-agent configurations and across backbone scales. We trace its underlying causes to three factors: the semantic limitations of condensed content, internal processing biases that suppress experience, and task regimes where pretrained priors already suffice. These findings challenge prevailing assumptions about self-evolving methods and underscore the need for more faithful and reliable approaches to experience integration.
title Large Language Model Agents Are Not Always Faithful Self-Evolvers
topic Computation and Language
url https://arxiv.org/abs/2601.22436