Paper page - Task-Focused Memorization for Multimodal Agents

Source: https://huggingface.co/papers/2605.31075

Abstract

A reinforcement-learning-based framework called TaskMem is introduced to dynamically determine what information to store in long-term memory for multimodal agents, improving performance on streaming video benchmarks.

Long-term memoryis essential formultimodal agentsto build coherent experience, accumulate world knowledge, and achievecontinual learning. However, constructing effective memory goes beyond memory module design and basic requirements such as accuracy and fidelity; the key challenge lies in determining what to memorize.Multimodal agents, such as embodied agents, continuously perceive, reason, and act in real or virtual environments, receiving an unbounded stream of multimodal observations. From this combinatorial explosion of information, an agent must selectively retain content that is relevant to its role in the environment and valuable for future tasks. To bridge this gap, we frame memory generation as a learnablememorization policyand introduce TaskMem (Task-focused MemorizationPolicy Learning), a reinforcement-learning-based framework that enables the policy to dynamically adjust its focus to the demands of real tasks encountered in the environment. TaskMem adopts atwo-phase trainingparadigm: Phase One learns how to memorize by optimizingmemory qualityunder fundamental fidelity requirements; Phase Two occurs after deployment, where the agent learns what to memorize by tuning an adapter on its base MLLM, using recent environment tasks to define areward modelthat guides thememorization policytoward task-relevant content. To evaluate our approach, we reformulateVideoMME,EgoLife, andEgoTempointostreaming benchmarksthat simulate a realistic setting in which an agent processes streaming observations and handles tasks arriving online. To isolate memory assessment, the questions must be answered using only the agent’s memory, without access to raw video. Built onQwen3-VL-30B-A3B, TaskMem improvesVQA accuracyby 6.3%, 7.0%, and 5.3% on these benchmarks, respectively.

View arXiv pageView PDFProject pageAdd to collection

Get this paper in your agent:

hf papers read 2605\.31075

Don’t have the latest CLI?curl \-LsSf https://hf\.co/cli/install\.sh \| bash

Models citing this paper0

No model linking this paper

Cite arxiv.org/abs/2605.31075 in a model README.md to link it from this page.

Datasets citing this paper0

No dataset linking this paper

Cite arxiv.org/abs/2605.31075 in a dataset README.md to link it from this page.

Spaces citing this paper0

No Space linking this paper

Cite arxiv.org/abs/2605.31075 in a Space README.md to link it from this page.

Collections including this paper0

No Collection including this paper

Add this paper to acollectionto link it from this page.