From Recall to Forgetting: Benchmarking Long-Term Memory for Personalized Agents

Md Nayem Uddin Kumar Shubham, Eduardo Blanco Chitta Baral Gengyu Wang, Arizona State University · 2026

Read From Recall to Forgetting: Benchmarking Long-Term Memory for Personalized Agents by Md Nayem Uddin Kumar Shubham, Eduardo Blanco Chitta Baral Gengyu Wang, Arizona State University (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2604-20006.txt.

Description

Uddin and four colleagues introduce Memora, a benchmark for long-term memory in personalized agents over weeks to months of chats. Past benchmarks mostly test fact retrieval, missing buildup over time and frequent updates. Memora covers remembering, reasoning, and recommending, with grounding checks and human review. A new metric, Forgetting-Aware Memory Accuracy, penalizes reliance on outdated memory, and tests show agents often reuse invalid memories despite small gains from memory agents.

Catalog record

CREATORMd Nayem Uddin Kumar Shubham, Eduardo Blanco Chitta Baral Gengyu Wang, Arizona State University
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE111045
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2604.20006

Browse the full library · About the Tower of Babel