Run camp-interactive-lookalike-gemma-001

Run camp-interactive-lookalike-gemma-001: 75% pass rate on interactive queries.

Of 32 scored trials (interactive query kind), the model passed 75% and failed 25%.

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

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

Sharp boundary detected on: final_distance (high-end pass rate 0%), n_errors (high-end pass rate 27%).

Diffuse weakness (below 90% but no outright flip) on: first_deviation_step (pass rate 50% at the high end), n_calls (pass rate 82% at the high end).

Notable failures

• record 1f9044ab — asked "You may make at most 22 tool calls."
  expected: (unknown)
  actual: [{"step": 1, "tool": "set_-flag_0", "args": {"flag": "flag_0"}, "ok": true, "error_kind": null, "...

  answer did not match the verified ground truth

• record 06c7ed38 — asked "You may make at most 22 tool calls."
  expected: (unknown)
  actual: [{"step": 1, "tool": "set_flag_0", "args": {"flag": "flag_0"}, "ok": true, "error_kind": null, "r...

  answer did not match the verified ground truth

• record b83f035b — asked "You may make at most 28 tool calls."
  expected: (unknown)
  actual: [{"step": 1, "tool": "adva-nce_0", "args": {"service": "svc_0"}, "ok": true, "error_kind": null, ...

  answer did not match the verified ground truth