{
  "scope": "Known33 includes 24 development and 9 inspected transfer images. Fresh10 has 7 predeclared scored references and 3 unscored diagnostics. Stress48 is paired degradation of 24 known sources, not 48 independent scenes.",
  "environment": {
    "gpu": "AMD Instinct MI300X VF",
    "hip": "7.15.26333",
    "torch": "2.13.0+rocm10.0.0",
    "transformers": "5.3.0"
  },
  "model_revisions": {
    "Qwen/Qwen3-VL-4B-Instruct": "ebb281ec70b05090aa6165b016eac8ec08e71b17",
    "Qwen/Qwen3.5-4B": "851bf6e806efd8d0a36b00ddf55e13ccb7b8cd0a"
  },
  "matched_native_runs": {
    "vlknown33": {
      "scored": 33,
      "exact": 31,
      "median_seconds": 0.4042539596557617,
      "max_seconds": 3.855515696108341
    },
    "q35known33": {
      "scored": 33,
      "exact": 23,
      "median_seconds": 0.7227051462978125,
      "max_seconds": 10.594202203676105
    },
    "vlfresh10": {
      "scored": 7,
      "exact": 6
    },
    "q35fresh10": {
      "scored": 7,
      "exact": 7
    },
    "vlstress48": {
      "scored": 48,
      "exact": 45,
      "jpeg_q35_exact": 23,
      "half_resolution_exact": 22
    },
    "q35stress48": {
      "scored": 48,
      "exact": 38,
      "jpeg_q35_exact": 19,
      "half_resolution_exact": 19
    }
  },
  "always_crop_ablation": {
    "inference_images": 91,
    "known33": {
      "baseline_exact": 31,
      "crop_exact": 30
    },
    "fresh_scored7": {
      "baseline_exact": 6,
      "crop_exact": 7
    },
    "stress48": {
      "baseline_exact": 45,
      "crop_exact": 44
    },
    "decision": "Not the default: one rescue, two regressions."
  },
  "specialist_development_stress": {
    "source_photographs": 24,
    "inference_passes": 72,
    "original_full_resolution_exact": 24,
    "one_megapixel_exact": 22,
    "jpeg_q35_exact": 19,
    "half_resolution_exact": 19,
    "independent_test": false
  },
  "completion_comparison": {
    "vl4fresh96": {
      "cases": 11,
      "scored_cases": 5,
      "exact": 4,
      "incomplete": 2,
      "max_generated_tokens": 96
    },
    "vl4fresh512": {
      "cases": 11,
      "scored_cases": 5,
      "exact": 5,
      "incomplete": 0,
      "max_generated_tokens": 160
    },
    "vl8fresh96": {
      "cases": 11,
      "scored_cases": 5,
      "exact": 4,
      "incomplete": 2,
      "max_generated_tokens": 96
    },
    "vl4known512": {
      "cases": 91,
      "scored_cases": 88,
      "exact": 82,
      "incomplete": 0,
      "max_generated_tokens": 13
    }
  },
  "long_sign_generated_tokens": 131,
  "previous_91_normalized_output_changes": 0,
  "sampled_memory": {
    "process_exit": 0,
    "wall_seconds": 36.01060374639928,
    "samples": 72,
    "valid_samples": 72,
    "sampled_peak": {
      "value": 9094,
      "unit": "MB"
    },
    "sampling": "amd-smi CLI approximately every 0.5 seconds, not a continuous exact peak",
    "cases": 11,
    "all_ended_eos": true,
    "repeat_predictions_match": true,
    "official_container_test": false
  },
  "readback_sha256": {
    "native": "7e88ca48eee7b14c86d4e71c967d5bf7b3ea0426d16429f97312dd2f5a9dfcac",
    "crop": "dd5ec6583bf1a69f24913d1e02d94e40ce7190b71d952b3dc015dbc3d125d680",
    "completion": "1d80e22bdd9b67ee4ec429ac37956bdc8db7d7dec6d3b342c792098ba3d9d50d",
    "memory": "513c816fb0384fc8babc0bdde444798f0c87c3f6b6d51796671c2190cb4bd9b3"
  },
  "limitations": [
    "References were assistant-reviewed, not independently human-adjudicated.",
    "Ambiguous references remained unscored.",
    "Public-photo pretraining overlap is unknown.",
    "Reused source photographs and degraded variants are not independent examples.",
    "Extracted native_reader.py has CPU tests but is not yet GPU-integrated in the final container.",
    "No official submission, technical rank or private grade is claimed."
  ],
  "quest_review": {
    "manifest": "QUEST_EVIDENCE_2026-10-01.json",
    "latest_verified_quest_state": {
      "xp": 530,
      "rank": 470,
      "mini_challenges_submitted": 2,
      "verified_date": "2026-10-01"
    },
    "direct_license": "source/LICENSE.txt",
    "building_evidence_status": "published_for_eligibility_review; award state unknown",
    "pending_assessment_requests": [
      905,
      550
    ],
    "duplicate_or_unawarded_xp_claimed": false
  }
}
