Story
arxiv_llm_reliability ยท Nov 2, 2026 ยท paper
Source brief
What Is This Memory Useful For? Measuring Task-Conditioned Utility in Long-Term Agent Memory
arxiv.orgNov 2, 2026
original source linked
In brief
A stored experience may answer one future question and offer little help with another. We measure this task-conditioned utility rather than assigning each memory a single importance score. A task-blind writer stores a...
Feed lens
agenteval