{
  "source_commit": "be175a0ae84bc92cdf8b703ac9660386c4e56a79",
  "selection": "Latest fully successful committed 44-probe run per model; archived responses only, no new model calls.",
  "files": {
    "haiku-4-5": "probe/results/claude-haiku-4-5__2026-04-18-b10f7ecb66a5__20260418T193004Z.jsonl",
    "sonnet-4-6": "probe/results/claude-sonnet-4-6__2026-04-18-b10f7ecb66a5__20260418T200003Z.jsonl",
    "opus-4-7": "probe/results/claude-opus-4-7__2026-04-18-b10f7ecb66a5__20260418T195958Z.jsonl"
  },
  "comparisons": [
    {
      "pair": [
        "haiku-4-5",
        "sonnet-4-6"
      ],
      "category_mean": 0.878,
      "categories": {
        "capability": {
          "compared": 7,
          "flagged": 1,
          "rate": 0.143
        },
        "identity": {
          "compared": 6,
          "flagged": 6,
          "rate": 1.0
        },
        "persona": {
          "compared": 6,
          "flagged": 6,
          "rate": 1.0
        },
        "refusal": {
          "compared": 8,
          "flagged": 8,
          "rate": 1.0
        },
        "safety_adjacent": {
          "compared": 6,
          "flagged": 6,
          "rate": 1.0
        },
        "style": {
          "compared": 7,
          "flagged": 7,
          "rate": 1.0
        },
        "tool_calling": {
          "compared": 4,
          "flagged": 4,
          "rate": 1.0
        }
      },
      "flag_types": {
        "response_changed": 25,
        "identity_change": 8,
        "length_delta": 5
      }
    },
    {
      "pair": [
        "haiku-4-5",
        "opus-4-7"
      ],
      "category_mean": 0.792,
      "categories": {
        "capability": {
          "compared": 7,
          "flagged": 0,
          "rate": 0.0
        },
        "identity": {
          "compared": 6,
          "flagged": 6,
          "rate": 1.0
        },
        "persona": {
          "compared": 6,
          "flagged": 5,
          "rate": 0.833
        },
        "refusal": {
          "compared": 8,
          "flagged": 8,
          "rate": 1.0
        },
        "safety_adjacent": {
          "compared": 6,
          "flagged": 6,
          "rate": 1.0
        },
        "style": {
          "compared": 7,
          "flagged": 5,
          "rate": 0.714
        },
        "tool_calling": {
          "compared": 4,
          "flagged": 4,
          "rate": 1.0
        }
      },
      "flag_types": {
        "identity_change": 9,
        "length_delta": 2,
        "response_changed": 23
      }
    },
    {
      "pair": [
        "sonnet-4-6",
        "opus-4-7"
      ],
      "category_mean": 0.798,
      "categories": {
        "capability": {
          "compared": 7,
          "flagged": 1,
          "rate": 0.143
        },
        "identity": {
          "compared": 6,
          "flagged": 6,
          "rate": 1.0
        },
        "persona": {
          "compared": 6,
          "flagged": 5,
          "rate": 0.833
        },
        "refusal": {
          "compared": 8,
          "flagged": 8,
          "rate": 1.0
        },
        "safety_adjacent": {
          "compared": 6,
          "flagged": 6,
          "rate": 1.0
        },
        "style": {
          "compared": 7,
          "flagged": 6,
          "rate": 0.857
        },
        "tool_calling": {
          "compared": 4,
          "flagged": 3,
          "rate": 0.75
        }
      },
      "flag_types": {
        "length_delta": 2,
        "identity_change": 11,
        "response_changed": 22
      }
    }
  ]
}
