Evidence and limits
Finding: The publisher record reports that decomposing answers into claims produced a useful uncertainty signal, while raw estimates still required post-hoc calibration.
Impact: Watch claim-level evidence and calibration as a candidate measurement approach for the recorded review topic (see source) and the recorded review topic (see source).
Boundary: One small benchmark, one answer and verifier model family, automated judging, and no human participants do not establish human trust, independent verification, or kit gains. Next: does the signal remain calibrated across models, datasets, evidence sources, and human review decisions? Candidate; not promoted or adopted
Flags: small single-benchmark evaluation; fixed calibration and held-out split; one model family; no human participants; closest calibrated baseline differences require caution; self-verification is not independent evidence
Record State: Candidate only: not accepted into the reviewed catalog
What was reviewed: Previously screened candidate; review remains limited