Help and Security
Evaluation troubleshooting
Diagnose agent contracts, credentials, queues, timeouts, artifacts, and cancellation.
Check the latest evaluation before starting another one:
Common causes are an invalid agent contract, a missing provider credential, an
unsupported dataset role, an active evaluation for the same experiment, a
timeout, or a run-time agent failure. Read the status and logs to distinguish
these cases.
Cancel only the evaluation you intend to stop:
Some private datasets return redacted diagnostics by design. Use the result and
allowed artifacts to create the next hypothesis instead of trying to retrieve
withheld detail.