Newest papers

Uni-Mem: Learning a Long-Horizon Agentic Memory Manager with Unified Online Policy Optimization

Zhiqian Xia, Jielong Tang, Ling-An Zeng, Shixiong Kai, Rong Yao, Hao Kong, Xiao Dong, Wei Wang · 2026 · arXiv

DOI: 10.48550/arXiv.2610.29671 · open-access PDF

Spanish

La memoria de agentes a largo plazo está cambiando de flujos de trabajo predefinidos a políticas aprendibles para la construcción y recuperación de memoria. Los métodos existentes basados en el aprendizaje optimizan la construcción por sí solos o combinan el aprendizaje de refuerzo en línea (RL) del lado de la recuperación con la supervisión de la construcción fuera de línea, limitando la coadaptación entre las dos etapas. En este artículo, exploramos la RL en línea de fin a fin para un gestor de memoria unificado (Uni-Mem) que aprende ambas capacidades bajo una política compartida. Sin embargo, la optimización conjunta con los algoritmos de RL existentes como GRPO se enfrenta a comparaciones de grupos confusas y feedba escasa.

Machine translation (Meta M2M100); the English original follows.

Abstract

Long-term agent memory is shifting from predefined workflows toward learnable policies for memory construction and retrieval. Existing learning-based methods either optimize construction alone or combine retrieval-side online reinforcement learning (RL) with offline construction supervision, limiting co-adaptation between the two stages. In this paper, we explore end-to-end online RL for a unified memory manager (Uni-Mem) that learns both capabilities under a shared policy. However, joint optimization with existing RL algorithms such as GRPO faces confounded group comparisons and sparse feedba

Read in: English · Spanish · Chinese · Hindi · Arabic · Portuguese · French · German · Japanese · Russian · Korean · Indonesian · Vietnamese

Is this paper real? This page was built from OpenAlex and arXiv records, not generated. Software can verify any reference in one call: GET https://store.agentexchange.work/gw/papers/verify?title=...&year=... (0.003 USD), fetch this record as JSON with /gw/papers/paper?id=2610.29671, or list what it cites with /gw/papers/references?id=.... Pay per call in USDC (x402) or with a 19 USD card key for 1,000 calls. Catalogue · llms.txt

Sources: OpenAlex (CC0), arXiv, Crossref. Agent Exchange