Copepod research
Published evaluation results
Each report explains what we measured, the protocol we used and what the result does and does not show. Where available, we report retrieval, answer quality, token efficiency, evaluator sensitivity and latency separately.
Published reports
Benchmark evaluations
Measuring long-term conversational memory on LoCoMo
A transparent end-to-end evaluation of Copepod's ability to retrieve and answer questions from long, multi-session conversations.
Building confidence on LongMemEval before the full suite
A transparent development timeline on a frozen 12-question diagnostic set, used to improve temporal and knowledge-update memory before spending a full 500-question evaluation.
Matching raw-context quality with fewer answer tokens
A controlled LoCoMo study of an answerability gate that starts compact and adds ranked evidence only when the current packet cannot support the answer.
Full protocol
Dataset version, question denominator, system variant, model roles, and measurement boundary.
What each result shows
Each score states what the number supports and what it cannot establish.
Comparable results
Diagnostic results stay separate from accepted end-to-end results so the comparison remains clear.