{
  "frozen_at": "2026-07-18T23:46:57.951556+00:00",
  "git_sha": "f93f4b5578c221d3d2665f9b305f3e86759be862+dirty",
  "spec": {
    "analysis": "campaign.evalaware.analyze",
    "hypotheses": [
      {
        "id": "H-aware-probe",
        "prediction": {
          "ci_excludes": null,
          "comparator": ">",
          "effect": null,
          "metric": "real_probe_balanced_acc",
          "threshold": 0.6
        },
        "scoreboard_row": null,
        "statement": "a benchmark-versus-organic direction is decodable"
      },
      {
        "id": "H-aware-steer",
        "prediction": {
          "ci_excludes": null,
          "comparator": ">",
          "effect": null,
          "metric": "delta_r_per_steer",
          "threshold": 0.05
        },
        "scoreboard_row": null,
        "statement": "steering along the direction inflates the score"
      },
      {
        "id": "H-aware-organic-fpr",
        "prediction": {
          "ci_excludes": null,
          "comparator": "<",
          "effect": null,
          "metric": "organic_fpr",
          "threshold": 0.1
        },
        "scoreboard_row": null,
        "statement": "the probe does not spuriously flag organic prompts (equivalence bound, an affirmative null)"
      }
    ],
    "id": "campaign-eval-aware",
    "kill_criteria": [
      {
        "comparator": "<",
        "id": "K-aware",
        "metric": "real_probe_balanced_acc",
        "threshold": 0.55
      }
    ],
    "science": "S16-robustness",
    "subjects": {
      "datasets": [
        "evalaware-paired",
        "wildchat-organic",
        "arena-organic"
      ],
      "extra": {
        "slice_hashes": {
          "arena-organic": "ds:a59032da6e2ab4e7fa4ad2cc647f4734",
          "evalaware-paired": "ds:94c82e6925d2654c1bb974be5e76b5ee",
          "wildchat-organic": "ds:e57a0be8be68113371133cea828e6e8b"
        }
      },
      "organisms": [],
      "signals": [
        "evalaware-rms"
      ]
    },
    "title": "The RM knows it is being tested, and steering along that direction inflates reward",
    "version": 1
  },
  "spec_hash": "spec:3671c68e023554a042cd340ae8dded8e",
  "study_id": "study:campaign-eval-aware@v1#3671c68e"
}