Controllable Memory Usage: Balancing Anchoring and Innovation in Long-Term Human-Agent Interaction

Fuente: arXiv
Saved in:
Bibliographic Details
Main Authors: Tian, Muzhao, Huang, Zisu, Wang, Xiaohua, Xu, Jingwen, Guo, Zhengkang, Qian, Qi, Shen, Yuanzhe, Song, Kaitao, Yuan, Jiakang, Lv, Changze, Zheng, Xiaoqing
Format: Preprint
Published: 2026
Subjects:
Online Access:
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1866915716955897856
author Tian, Muzhao
Huang, Zisu
Wang, Xiaohua
Xu, Jingwen
Guo, Zhengkang
Qian, Qi
Shen, Yuanzhe
Song, Kaitao
Yuan, Jiakang
Lv, Changze
Zheng, Xiaoqing
author_facet Tian, Muzhao
Huang, Zisu
Wang, Xiaohua
Xu, Jingwen
Guo, Zhengkang
Qian, Qi
Shen, Yuanzhe
Song, Kaitao
Yuan, Jiakang
Lv, Changze
Zheng, Xiaoqing
contents As LLM-based agents are increasingly used in long-term interactions, cumulative memory is critical for enabling personalization and maintaining stylistic consistency. However, most existing systems adopt an ``all-or-nothing'' approach to memory usage: incorporating all relevant past information can lead to \textit{Memory Anchoring}, where the agent is trapped by past interactions, while excluding memory entirely results in under-utilization and the loss of important interaction history. We show that an agent's reliance on memory can be modeled as an explicit and user-controllable dimension. We first introduce a behavioral metric of memory dependence to quantify the influence of past interactions on current outputs. We then propose \textbf{Stee}rable \textbf{M}emory Agent, \texttt{SteeM}, a framework that allows users to dynamically regulate memory reliance, ranging from a fresh-start mode that promotes innovation to a high-fidelity mode that closely follows interaction history. Experiments across different scenarios demonstrate that our approach consistently outperforms conventional prompting and rigid memory masking strategies, yielding a more nuanced and effective control for personalized human-agent collaboration.
format Preprint
id arxiv_https___arxiv_org_abs_2601_05107
institution arXiv
publishDate 2026
record_format arxiv
spellingShingle Controllable Memory Usage: Balancing Anchoring and Innovation in Long-Term Human-Agent Interaction
Tian, Muzhao
Huang, Zisu
Wang, Xiaohua
Xu, Jingwen
Guo, Zhengkang
Qian, Qi
Shen, Yuanzhe
Song, Kaitao
Yuan, Jiakang
Lv, Changze
Zheng, Xiaoqing
Artificial Intelligence
As LLM-based agents are increasingly used in long-term interactions, cumulative memory is critical for enabling personalization and maintaining stylistic consistency. However, most existing systems adopt an ``all-or-nothing'' approach to memory usage: incorporating all relevant past information can lead to \textit{Memory Anchoring}, where the agent is trapped by past interactions, while excluding memory entirely results in under-utilization and the loss of important interaction history. We show that an agent's reliance on memory can be modeled as an explicit and user-controllable dimension. We first introduce a behavioral metric of memory dependence to quantify the influence of past interactions on current outputs. We then propose \textbf{Stee}rable \textbf{M}emory Agent, \texttt{SteeM}, a framework that allows users to dynamically regulate memory reliance, ranging from a fresh-start mode that promotes innovation to a high-fidelity mode that closely follows interaction history. Experiments across different scenarios demonstrate that our approach consistently outperforms conventional prompting and rigid memory masking strategies, yielding a more nuanced and effective control for personalized human-agent collaboration.
title Controllable Memory Usage: Balancing Anchoring and Innovation in Long-Term Human-Agent Interaction
topic Artificial Intelligence
url https://arxiv.org/abs/2601.05107