{
  "date": "2026-10-15",
  "platform": "Linux x86-64",
  "python": "3.11.9",
  "status": "integration-checked",
  "release_eligible": false,
  "tests": {
    "passed": 73,
    "inspect_installed": true,
    "external_model_api_calls": 0
  },
  "training_packages": {
    "trl": "0.29.0",
    "transformers": "5.3.0",
    "datasets": "4.6.1",
    "accelerate": "1.12.0",
    "peft": "0.18.1",
    "torch": "2.10.0+cpu"
  },
  "sft": {
    "status": "completed",
    "steps": 1,
    "loss": 2.721248149871826,
    "grad_norm": 3.143519639968872,
    "parameter_change_assertion": "passed"
  },
  "grpo": {
    "status": "completed",
    "steps": 1,
    "loss": 0.0,
    "grad_norm": 1.5283790826797485,
    "reward_mean": 0.03125,
    "reward_std": 0.1767766922712326,
    "parameter_change_assertion": "passed"
  },
  "real_cpu_evaluation": {
    "smoke_only": true,
    "release_eligible": false,
    "compatibility": {
      "passed": true,
      "release_eligible": false,
      "reasons": [],
      "baseline_accuracy": 0.0,
      "candidate_accuracy": 0.0,
      "delta": 0.0,
      "max_drop": 1.0,
      "category_deltas": {
        "arithmetic": 0.0,
        "constraints": 0.0,
        "logic": 0.0,
        "units": 0.0
      },
      "note": "Descriptive regression gate only; inspect category-level outcomes and apply an independently approved release policy."
    },
    "notice": "Real CPU generation with tiny synthetic checkpoints. Full regression tolerance tests artifact compatibility only; no quality or training improvement claim."
  },
  "fixture_regression_comparison": {
    "passed": true,
    "release_eligible": false,
    "reasons": [],
    "baseline_accuracy": 0.5,
    "candidate_accuracy": 1.0,
    "delta": 0.5,
    "max_drop": 0.0,
    "category_deltas": {
      "arithmetic": 1.0,
      "constraints": 0.0,
      "logic": 1.0,
      "units": 0.0
    },
    "note": "Descriptive regression gate only; inspect category-level outcomes and apply an independently approved release policy."
  },
  "local_release_bundle": {
    "status": "candidate",
    "published": false,
    "release_eligible": false,
    "files": 10
  },
  "agent_lightning": {
    "revision": "69f5a40b36d9bff646b8e7a5532e8b78bd41eaf9",
    "source_check": "passed against real checkout",
    "tool_loop_tests": 4,
    "gpu_execution": "not run"
  },
  "source_sha256": {
    "examples/agent_harness.py": "0c852a82e1c85e3ebc641bf9c79e1572c355212868c4e78d5f7cbf6a0556e4c8",
    "examples/agent_lightning/hessian_lightning_harness.py": "bbdec718f9e1fd777bc88308ac972ea464ad501434dd84a260f7eb39b977db6c",
    "examples/agent_lightning/launch.py": "a05913ebb73ffa52ca77e06814d6be01051746b4cd7bc8e304a44cf896a2ee41",
    "examples/agent_lightning/train.py": "9a74ba25cad340ecd730f86b56e92803358a0d5670f1e95b9efd256d7a845d32",
    "examples/evaluation_cpu_smoke.py": "5cb87036d41041c2d7d13516b7beffeb8b3d916c086e8d64d73631b1547204fd",
    "examples/training/cpu_smoke.py": "d8cfac90d854a0fb198588255c401b544bb4d0d4eea24cdbc3f5fc97d35d2b6e",
    "src/hessian_recipes/__init__.py": "789c960cdc4eca10e9b04ed4a66bdcb59f4c7236778ae6ba139277fa8c655ac7",
    "src/hessian_recipes/__main__.py": "935a1c1166b0c1ea35a82256345000bf2c73ded718d77773bc27a71ecce28f7d",
    "src/hessian_recipes/cli.py": "c1408af82d3c5f69625f8c73621c04bda3139ea35e8a3cda4ae930d66f06524e",
    "src/hessian_recipes/distillation.py": "5c395accd85980afbf493391fac6f819624385c021c9ae5c874be36d350000cd",
    "src/hessian_recipes/evaluation.py": "90025bf305339d346cf2889e7c443c345fadfbc1055a08de40f2d7ba35288d8a",
    "src/hessian_recipes/inspect_tasks.py": "7a02b2ac0eecb997c079b78b6b067d9f0576f9cb266a832821f430c6d092e64c",
    "src/hessian_recipes/releases.py": "2a402248e9a092614bf38e022002d661c6adc89c07c092c884a7efa8d0c43d5d",
    "src/hessian_recipes/training.py": "a00a54cc671bceed4a34ca3710fcf8fa5faf7c806e55ef7fbb5b5d6448253f81",
    "src/hessian_recipes/trajectories.py": "58444e77707284835840f3766f3faa1034d6a8e1dc08705b288690e7e9b6892f"
  },
  "not_validated": [
    "model quality",
    "GPU execution",
    "cluster/container runtime",
    "distributed training",
    "real Lightning optimizer update",
    "domain models",
    "hosted service",
    "Hub publication"
  ],
  "source_commit": "8bc44ee7451bcf08b7ffe53d2582530e97f6c3c2",
  "scope": "Initial foundation; German Qwen GPU evidence is recorded separately."
}