Runs, missing outputs, and output mapping
Symptom
An experiment cannot start, has incomplete behavior, or contains unmatched external responses.
Likely causes
| What you observe | Focused guide |
|---|---|
| Managed execution or evaluation progress is blocked | Benchmark run troubleshooting |
| A response or recorded failure is absent | Missing Outputs |
| Imported rows do not match the exact Case manifest | Output mapping troubleshooting |
Fix
Identify the Run, candidate, Benchmark Version, and execution path. Separate infrastructure failure, candidate failure, missing submission, mapping rejection, and evaluation completeness before interpreting scores.
Follow the focused guide for the actual state you observe; keep accepted records and explicit human decisions intact.
Prevention
Preserve exact object and version identities, review the relevant source context, and inspect the accepted state before comparing outcomes.