DocsTroubleshootingEvaluation Errors

Evaluation Errors

Error categories, retryable vs non-retryable, and how to retry.

Candidate generation, deterministic validation, and judging are separate parts of a Monitor Run. The report identifies which part failed; missing telemetry is never counted as a model loss.

CategoryMeaningLikely action
Provider or rate-limit errorTemporary upstream failureWait for the worker retry or start again after the provider recovers
TimeoutA call exceeded its execution limitCheck provider health and prompt size
Model unavailableThe catalog model cannot currently executeChoose another candidate or wait for catalog sync
Context too longThe rendered request exceeds model capabilityShorten the input or choose a larger-context model
Content filter or refusalThe provider did not return a usable answerInspect the item before changing the prompt or candidate
Judge parse failureA judge did not return the pairwise contractPeerLM records the failed judgment; the panel may make the item indeterminate
Validator failureA declarative requirement failedInspect the exact text, JSON path, schema, or regex result

Judges return a winner, failure tag, and one sentence. There is no per-evaluator 1–10 score to retry or recompute.