{
  "study": "glm-5.2-frontier-256k-siliconflow-v1",
  "status": "Reliability gate not met",
  "claim_status": "Not supported",
  "model": "z-ai/glm-5.2",
  "provider": "SiliconFlow",
  "provider_pinned": true,
  "provider_mismatches": 0,
  "conditions": {
    "prompt_version": 1,
    "variables": 20,
    "max_completion_tokens": 262144,
    "temperature": 0.6,
    "reasoning_effort": "high",
    "max_calls": 10,
    "max_spend_usd": null,
    "difficulties": [4.2, 4.8, 5.4, 6.0, 6.6]
  },
  "counts": {
    "calls": 10,
    "correct_completed": 3,
    "silently_wrong_completed": 3,
    "loud_failure": 4
  },
  "loud_failures": {
    "truncation": 0,
    "parse_failure": 1,
    "api_error": 3
  },
  "levels": {
    "4.2": {"calls": 2, "correct": 1, "silent_wrong": 0, "loud": 1},
    "4.8": {"calls": 2, "correct": 0, "silent_wrong": 1, "loud": 1},
    "5.4": {"calls": 2, "correct": 0, "silent_wrong": 0, "loud": 2},
    "6.0": {"calls": 2, "correct": 1, "silent_wrong": 1, "loud": 0},
    "6.6": {"calls": 2, "correct": 1, "silent_wrong": 1, "loud": 0}
  },
  "paired_summary": {
    "scout_32768": {"correct": 4, "silent_wrong": 0, "loud": 6},
    "followup_262144": {"correct": 3, "silent_wrong": 3, "loud": 4},
    "truncations": {"scout_32768": 4, "followup_262144": 0},
    "api_errors": {"scout_32768": 1, "followup_262144": 3},
    "interpretation": "The provider and token cap both changed, so paired differences cannot be attributed to the cap alone."
  },
  "gates": {
    "no_more_than_two_loud_failures": false,
    "at_least_two_correct_completed": true,
    "at_least_two_silently_wrong_completed": true,
    "reliability_pass": false,
    "completed_answer_mix_available": true
  },
  "scoring_correction": {
    "reason": "The original parser rejected contiguous 20-bit FINAL assignments.",
    "changed_calls": [
      {"call_id": 12, "before": "fail_parse", "after": "pass"},
      {"call_id": 18, "before": "fail_parse", "after": "pass"}
    ],
    "genuine_19_bit_answer_remained": "fail_parse"
  },
  "spend": {
    "actual_usd": 0.66254,
    "configured_ceiling_usd": null
  },
  "decision": "Report the failed reliability gate. Do not fit or claim a validated confidence model from six completed answers; any larger focus study requires a new protocol.",
  "provenance": {
    "collection_commit": "acef866",
    "scoring_commit": "4a1c50c",
    "config_sha256": "4f0c38170ffd1ad073d7ac9b4efd46abd1de5c8366337bb3d976a35d32bf0680",
    "database_sha256_after_rescore": "0f7afacca24605b17d786146f825135a36e348b5e41b13f0bbafe3bc3b7525d7",
    "compact_export_sha256": "b7567f1fba39e8e1d6e1a6de437d04eea96ba61b32a4912fdf7d64632e4ce066",
    "trace_export_sha256": "e472e739347cc767fd25126d2024be84445744dec2415e53cfa5249ee368b7ba"
  }
}