Run camp-tool-depth-ultra-003

Run camp-tool-depth-ultra-003: 95% pass rate on tool queries.

Of 56 scored trials (tool query kind), the model passed 95% and failed 5%.

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

Diffuse weakness (below 90% but no outright flip) on: catalog_size (pass rate 86% at the high end), chain_length (pass rate 86% at the high end), id_collisions_on (pass rate 89% at the high end), realized_plan_depth (pass rate 86% at the high end).

Notable failures

• record c3190c93 — asked "You may use at most 32 tool calls."
  expected: (unknown)
  actual: ]

  answer did not match the verified ground truth

• record 9fe6a4d9 — asked "You may use at most 32 tool calls."
  expected: (unknown)
  actual: ]

  answer did not match the verified ground truth

• record 4a2e0ba7 — asked "You may use at most 36 tool calls."
  expected: (unknown)
  actual: ]

  answer did not match the verified ground truth