Skip to main content

Observe and debug runs

Use run artifacts to understand quality and failure modes.

What to inspect first

  • Overall run status (queued, running, completed, failed)
  • Sample-level output quality and scorer deltas
  • Event timeline and trace artifacts

Common debugging flow

  1. Identify failing or low-score samples.
  2. Open sample artifacts and event logs.
  3. Adjust solver/scorer config.
  4. Re-run and compare results.

Reliability tip

Keep one stable reference suite/dataset for regression checks while iterating on new solver changes.