MemPrism：面向长程智能体的任务条件关系记忆视图

Authors:Bingfan Zeng, Zhisheng Chen, Zhengwei Xie, Bangde Cao, Yuxuan Li, Jinhan Li, Zheng Lu, Xiangchen Guan, Zikai Xiao, Rui Qian, Jingwei Song

View PDF HTML (experimental)

Abstract:Long-horizon agents performing multi-step Web tasks must reuse earlier page observations and interactions to support subsequent actions. However, preserving relevant history does not ensure that it exposes the relations needed for the current decision. We call this gap representation mismatch. To this end, we propose MemPrism, a task-conditioned relational memory framework that separates persistent experience storage from decision-time working memory. MemPrism records interactions as the event stream and dynamically constructs relational views according to the current task context. A lightweight view policy selects the relation structure, evidence range, outcome condition, and granularity, while a deterministic composer and renderer transform historical facts into a temporary optical working-memory view for a frozen task policy. Experiments on long-horizon embodied and web-agent benchmarks show that MemPrism consistently improves the task performance, especially as trajectories become longer, while reducing memory token consumption. Furthermore, the learned view policy transfers across different VLMs without additional adaptation, demonstrating the effectiveness of task-conditioned relational views as a general memory interface for agents.

Subjects:	Artificial Intelligence (cs.AI)
Cite as:	arXiv:2608.06745 [cs.AI]
	(or arXiv:2608.06745v2 [cs.AI] for this version)
	https://doi.org/10.48550/arXiv.2608.06745 arXiv-issued DOI via DataCite

Submission history

From: Zhisheng Chen [view email]
[v1] Fri, 7 Aug 2026 03:13:43 UTC (6,627 KB)
[v2] Tue, 27 Oct 2026 10:39:17 UTC (4,986 KB)