{
  "schema_version": 2,
  "manifest_ref": "agent-research:atomrec",
  "method": "atomrec-checkpoint",
  "dataset": {
    "name": "Amazon Beauty 2014 5-core",
    "users": 6,
    "candidate_count": 10,
    "split_policy": "per-user chronological validation/test; pre-target memory rebuilt for every case",
    "sha256": {
      "meta_Beauty.json.gz": "c7977dc0e0ead14ac7df8c1c6de74da00e4bebfad87a0f0c5763d4c5bc9b53a0",
      "reviews_Beauty_5.json.gz": "e5925ec99023f1dc9c7d3dff7ef34cfeacd40d1b981b5525e1e4ba9c8abc18fe"
    }
  },
  "seeds": [
    42,
    43,
    44
  ],
  "checkpoint": {
    "model_id": "Qwen/Qwen3-4B-Instruct-2507",
    "revision": "cdbee75f17c01a7cc42f958dc650907174af0554"
  },
  "validation": {
    "recent_history_hit_at_3": [
      0.5,
      0.5,
      0.6666666666666666
    ],
    "atomrec_hit_at_3": [
      0.3333333333333333,
      0.16666666666666666,
      0.3333333333333333
    ],
    "recent_history_ndcg_at_3": [
      0.37697658452381916,
      0.4384882922619096,
      0.46030991785715253
    ],
    "atomrec_ndcg_at_3": [
      0.25,
      0.10515495892857625,
      0.25
    ],
    "atomrec_valid_rate": [
      1.0,
      1.0,
      1.0
    ]
  },
  "test": {
    "recent_history_hit_at_3": [
      0.5,
      0.6666666666666666,
      0.6666666666666666
    ],
    "atomrec_hit_at_3": [
      0.0,
      0.0,
      0.0
    ],
    "recent_history_ndcg_at_3": [
      0.4384882922619096,
      0.5218216255952429,
      0.4384882922619096
    ],
    "atomrec_ndcg_at_3": [
      0.0,
      0.0,
      0.0
    ],
    "atomrec_valid_rate": [
      0.8333333333333334,
      0.8333333333333334,
      0.8333333333333334
    ]
  },
  "execution": {
    "generator_calls": 105,
    "memory_stages": [
      "atomic extraction",
      "semantic linking",
      "field evolution",
      "multi-hop retrieval",
      "grounded synthesis"
    ],
    "downstream_ranker": "frozen checkpoint mean-hidden-state cosine over supplied candidates",
    "failure_policy": "invalid generations score zero; no target or gold fallback"
  },
  "evaluation_protocol": {
    "tier": "l2_execution_evidence",
    "formal_comparison": false,
    "claim_policy": "small public checkpoint execution with a negative test result; not the paper benchmark",
    "seeds": [
      42,
      43,
      44
    ]
  },
  "provenance": {
    "artifact_path": "docs/agent-research/2609.04882-atomrec/metrics/amazon-beauty-checkpoint-seeds42-44.json",
    "dataset_fingerprint": "sha256:c7977dc0e0ead14ac7df8c1c6de74da00e4bebfad87a0f0c5763d4c5bc9b53a0+e5925ec99023f1dc9c7d3dff7ef34cfeacd40d1b981b5525e1e4ba9c8abc18fe"
  }
}