{
  "status": "supplemental_cpu_reverification_passed_pending_fresh_review",
  "date_utc": "2026-10-21T06:49:35.248308+00:00",
  "scope": "Independent supplement to an immutable real GPU probe; CPU-only replay of existing measured trace, no new inference or benchmark",
  "raw_report": "benchmarks/reports/phase2-hybrid-lagged-dev-qwen35-9b8.json",
  "raw_report_sha256": "51c4b6e7744129fcfe55c69644d6d6129f021f255987058a8427223161be2060",
  "source_archive": "benchmarks/reports/phase2-ds-source-v1-20261021/source.tar.gz",
  "source_archive_sha256": "f48d430d2433194f4680896e699760a8e0a9602505a320833c02fbbefabe60ab",
  "raw_summary_preserved": {
    "completed_candidate_requests": 1,
    "mature_core_decisions": 28,
    "all_outputs_exact_to_baseline": true,
    "failures": 0,
    "correctness_and_maturity_pass": false,
    "maturity_and_oracle_pass": false,
    "equivalent_speedup_claim_supported": false,
    "competitive_advantage_claim_supported": false,
    "functional_results_scope": "raw execution only; provenance checked separately after completion",
    "frozen_execution_valid": true,
    "accepted_frozen_probe_pass": false
  },
  "original_probe_accepted_flag_remains_false": true,
  "original_probe_verifier_error": "AssertionError: [{'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 31, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 32, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 33, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 34, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 35, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 36, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 37, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 38, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 39, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 40, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 41, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 42, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 43, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 44, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 45, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 46, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 47, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 48, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 49, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 50, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 51, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 52, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 53, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 54, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 55, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 56, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 57, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 58, 'reason': 'curve float32 projection'}]",
  "root_cause": "The independent verifier built curve as list-of-lists, while actual in-memory scheduler replay retained list-of-tuples. Exact container comparison raised curve float32 projection errors before C++ replay. JSON serialization changed tuple pairs to lists, so the unchanged original verifier independently passed the saved actual report.",
  "fix": "Normalize pair containers to lists only; all measured float values remain exact and altered values are still rejected.",
  "source_sha256": {
    "benchmarks/experiments/verify_lagged_trace.py": "f3b82f112bdc3348647a77ddee4348127ba7182ea9e1429690be5c4244c3b2b6",
    "tests/test_ds_experiment_harness.py": "ea63af29e2a929136c3bac0e0b7b4606e621e23eb058f6392848dab11614444b"
  },
  "original_archived_verifier_sha256": "dac5dfcd531db17a161abbc4dd84ba340e9b23b2779e33c4742160093846bd71",
  "original_archived_verifier_json_result": {
    "matches": true,
    "status": "verified",
    "decisions": 28,
    "streams": 1,
    "history_errors": [],
    "actual_seqlens_q": [
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4
    ],
    "expected_seqlens_q": [
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4
    ],
    "header_sha256": "164f45db4cccd6d6909dd2846e7dfffb57754b108af679443bf4bd71ec33785a",
    "performance_claim": false,
    "method": "independent chronological full-cost/age reconstruction and unchanged C++ MTPVarlen"
  },
  "original_archived_verifier_runtime_tuple_reproduction": {
    "raised": true,
    "error": "[{'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 31, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 32, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 33, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 34, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 35, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 36, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 37, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 38, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 39, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 40, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 41, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 42, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 43, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 44, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 45, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 46, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 47, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 48, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 49, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 50, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 51, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 52, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 53, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 54, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 55, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 56, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 57, 'reason': 'curve float32 projection'}, {'stream': 'qwen3.5-9b-8bit:nonthinking:0', 'cycle': 58, 'reason': 'curve float32 projection'}]"
  },
  "corrected_verifier_json_result": {
    "matches": true,
    "status": "verified",
    "decisions": 28,
    "streams": 1,
    "history_errors": [],
    "actual_seqlens_q": [
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4
    ],
    "expected_seqlens_q": [
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4
    ],
    "header_sha256": "164f45db4cccd6d6909dd2846e7dfffb57754b108af679443bf4bd71ec33785a",
    "performance_claim": false,
    "method": "independent chronological full-cost/age reconstruction and unchanged C++ MTPVarlen"
  },
  "corrected_verifier_runtime_tuple_result": {
    "matches": true,
    "status": "verified",
    "decisions": 28,
    "streams": 1,
    "history_errors": [],
    "actual_seqlens_q": [
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4
    ],
    "expected_seqlens_q": [
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      2,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4,
      4
    ],
    "header_sha256": "164f45db4cccd6d6909dd2846e7dfffb57754b108af679443bf4bd71ec33785a",
    "performance_claim": false,
    "method": "independent chronological full-cost/age reconstruction and unchanged C++ MTPVarlen"
  },
  "real_probe_observations": {
    "baseline_output_tokens": 128,
    "lagged_output_tokens": 128,
    "all_token_ids_exact": true,
    "completed_cycles": 58,
    "mature_core_decisions": 28
  },
  "reproduce_cpu_command": ".venv/bin/python benchmarks/experiments/verify_lagged_trace.py --input benchmarks/reports/phase2-hybrid-lagged-dev-qwen35-9b8.json",
  "regression_command": ".venv/bin/pytest -q tests/test_ds_experiment_harness.py",
  "gpu_executed_for_supplement": false,
  "performance_claim": false,
  "limits": [
    "The original report, source archive, failed in-probe verifier result and accepted=false summary are unchanged.",
    "This supplement verifies existing real-model output and mature algorithm decisions; it does not promote the original accepted flag or establish a speed gain.",
    "Future in-process probes need fresh reviewed source archives containing the schema correction."
  ]
}