MemBench: Towards More Comprehensive Evaluation on the Memory of LLM-based Agents

Catalog evidence and recorded chart appearances. Inclusion is not an endorsement.

Source snapshot imported 2026-10-20. This is not a live feed. Source and contributions.

Evidence and limits

Finding: Memory evaluation changes materially by memory level, interaction role, and resource dimension. One recall measure cannot represent the capability.

Impact: Strengthen multi-dimensional memory evaluation and participant-versus-observer distinctions. Reject one memory score.

Boundary: The benchmark does not evaluate governance, cross-context isolation, correction authority, or repository ownership. Next: which governance dimensions can be added without obscuring benchmark performance?

Flags: benchmark construct choices; no governance evaluation

Record State: Catalog review recorded; not a performance endorsement

What was reviewed: Structure and relevance screened; not a full evaluation

Inspect original source

Recorded chart appearances

  • 2026-10-13: position 40. Relevant to the recorded review topic (see source); Structure and relevance screened; not a full evaluation evidence
  • 2026-10-12: position 39. Relevant to the recorded review topic (see source); Structure and relevance screened; not a full evaluation evidence