UTILMEM: Benchmarking Evidence Utilization in Long-Term Conversational Memory
2026-08-31 • Computation and Language
Computation and Language
AI summaryⓘ
The authors explain that existing tests for chatbot memory mostly check if the chatbot can remember exact facts from past talks. However, real memory use is harder and involves putting together clues from long and sometimes confusing chats to complete tasks. They created a new test called UtilMem to see how well systems manage this kind of memory use, including reasoning, finding hidden relevant info, and ignoring distractions. Their results show that just finding information isn't enough; systems also need to combine it well and tell what's truly important. This highlights a big challenge in making chatbots with better long-term memory.
long-term memoryconversational agentsmemory utilizationfactual recallretrieval-based systemsmemory augmentationevidence integrationdistractor interferencebenchmarkdistributed evidence
Authors
Peijun Qing, Fobo Shi, Soroush Vosoughi
Abstract
Long-term memory is increasingly important for conversational agents, yet existing benchmarks primarily measure memory through pointwise factual recall: whether a system can recover isolated facts or event-level details from prior interactions. Real-world memory use, however, often requires a more demanding capability: integrating distributed, implicit, and noisy evidence across extended interaction histories into coherent, task-oriented outputs. We call this capability memory utilization. Here, we introduce UtilMem, a diagnostic benchmark comprising 1,717 instances across five domains, designed to evaluate four underexplored aspects of memory utilization: reasoning over dense histories, identifying implicitly relevant memories, synthesizing distributed evidence into summaries, analyses, or plans, and resisting interference from semantically similar distractors. Evaluating a diverse set of retrieval-based and memory-augmented systems, we find that strong performance on conventional factual-memory benchmarks does not reliably translate into effective memory utilization. Moreover, retrieval alone is insufficient: even when relevant evidence is successfully recovered, systems frequently fail to integrate information across sessions or to distinguish useful evidence from plausible distractors. These findings expose a substantial gap between accessing stored information and using it effectively, and suggest that progress in long-term conversational memory will require architectures that explicitly support evidence integration and robustness to retrieval interference. Code is available at https://github.com/peijunallin/UtilMem.