CloneMem: Benchmarking Long-Term Memory for AI Clones

TL;DR

CloneMem is a new benchmark designed to evaluate long-term memory in AI clones using real-world digital traces over multiple years, exposing current memory systems' limitations in modeling continuous personal histories.

Contribution

It introduces a novel, life-grounded benchmarking framework with hierarchical data construction for assessing long-term memory in AI clones, addressing limitations of existing fragmented benchmarks.

Findings

Current memory mechanisms perform poorly on CloneMem tasks.

The benchmark reveals significant challenges in modeling continuous personal histories.

CloneMem provides a new standard for evaluating long-term memory in personalized AI.

Abstract

AI Clones aim to simulate an individual's thoughts and behaviors to enable long-term, personalized interaction, placing stringent demands on memory systems to model experiences, emotions, and opinions over time. Existing memory benchmarks primarily rely on user-agent conversational histories, which are temporally fragmented and insufficient for capturing continuous life trajectories. We introduce CloneMem, a benchmark for evaluating longterm memory in AI Clone scenarios grounded in non-conversational digital traces, including diaries, social media posts, and emails, spanning one to three years. CloneMem adopts a hierarchical data construction framework to ensure longitudinal coherence and defines tasks that assess an agent's ability to track evolving personal states. Experiments show that current memory mechanisms struggle in this setting, highlighting open challenges for life-grounded personalized AI. Code and dataset are available at https://github.com/AvatarMemory/CloneMemBench