What it measures

Conversation-based tasks test an assistant’s ability to retain, update, and use information across an extended history.

How to read the result

Separate retrieval, temporal reasoning, and recognizing that an answer is unavailable. Remembered information can become stale or conflict with newer evidence.

Limits and comparison traps

Memory indexing, retrieval, and context assembly are part of the system. A long context window alone does not define the evaluated memory pipeline.

Published evaluation

End-to-end memory-question accuracy under the named history size and memory/reading setup. Native fractions are converted to percentages; oracle-session scores are not substituted.

Where this fits in an evaluation suite

A large context window is a capacity specification. Successful retrieval and cross-document reasoning are observed capabilities. Test both at the actual lengths your workflow needs.

Start with a capability hypothesis, then choose a second test with a different failure mode. Keep the checkpoint, reasoning effort, agent tools, sampling policy, and scoring procedure fixed when measuring a training intervention. Record failures and missing outputs in the denominator.

A reproducible comparison

End-to-end memory-question accuracy under the named history size and memory/reading setup. Native fractions are converted to percentages; oracle-session scores are not substituted.

Publish the exact run configuration, item coverage, score units, and available uncertainty. If a source supplies only an aggregate or a single run, preserve that limit. A new model name or a rounded score alone does not establish a capability improvement.

PRIMARY EVIDENCE & FURTHER READING

  1. LongMemEval: Benchmarking Chat Assistants on Long-Term Interactive Memory — published result table ↗

    Wu, Di, Wang, Hongwei, Yu, Wenhao, Zhang, Yuwei, Chang, Kai-Wei, Yu, Dong · 2024

  2. Primary benchmark documentation or model evaluation report ↗

Reviewed 2026-10-26. Explanation and interpretation by RLTasks. Benchmark ownership remains with its original authors.