Run camp-full_map-ultra-002/breaking_point

Run camp-full_map-ultra-002/breaking_point: passed essentially everything (98%) -- this test did not challenge the model.

Of 55 scored trials (belief query kind), the model passed 98%. That's saturation: the difficulty range probed here isn't finding this model's limits. Consider a harder preset (breaking_point or adversarial_belief) or a wider dimension range.

1 additional trial(s) ended in ERROR (infra failures, cost cap, or cancellation) and are excluded from the pass rate above.

1 failure(s) recorded; failures not yet verified -- use "Verify failures" on this page (or scripts/verify_past_failures.py).

1 belief failure(s) were classified against known shortcuts, but none of them dominate -- the failures don't share one obvious wrong-answer pattern.

Notable failures

• record d392ecae — asked "Where does Juno believe the narrow folder is? Answer with only the location name (or 'unknown' if they have no belief)."
  expected: the server room
  actual: meditation room

  the model's answer matches no known shortcut -- an unclassified miss