Run camp-full_map-super-002/breaking_point

Run camp-full_map-super-002/breaking_point: 96% pass rate on belief queries.

Of 56 scored trials (belief query kind), the model passed 96% and failed 4%.

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

In 1 of 2 classifiable belief failures the model's wrong answer matches the 'most_frequent_location' shortcut -- a plausible but ungrounded guess rather than tracking what that entity actually witnessed.

Notable failures

• record 26f5b058 — asked "Where does Quinn believe the bright basket is? Answer with only the location name (or 'unknown' if they have no belief)."
  expected: the basement
  actual: loading dock

  the model's answer matches the 'most_frequent_location' shortcut, not the verified belief

• record 1b2f79d9 — asked "Where does Marcus believe the shiny box is? Answer with only the location name (or 'unknown' if they have no belief)."
  expected: the wine cellar
  actual: reception area

  the model reported the object's TRUE final location, not what the queried entity actually believed -- it collapsed belief into reality