{
  "schema_version": 2,
  "method": "autolr",
  "dataset": "MovieLens 100K",
  "seeds": [
    42,
    43,
    44
  ],
  "runs": [
    {
      "paper": {
        "arxiv_id": "2609.04871",
        "title": "AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems",
        "url": "https://arxiv.org/abs/2609.04871",
        "organization": "NetEase, Inc."
      },
      "dataset": {
        "name": "MovieLens 100K",
        "users": 220,
        "items": 360
      },
      "setup": {
        "adapter": "autolr",
        "same_split_and_candidates": true,
        "seed": 42
      },
      "baseline": {
        "name": "transition + content + popularity",
        "hit_at_10": 0.10909090909090909,
        "ndcg_at_10": 0.05400546778896325,
        "fresh_hit_at_10": 0.031746031746031744,
        "head_share_at_10": 0.23772727272727273
      },
      "method": {
        "name": "autolr executable offline mechanism",
        "hit_at_10": 0.10909090909090909,
        "ndcg_at_10": 0.05400546778896325,
        "fresh_hit_at_10": 0.031746031746031744,
        "head_share_at_10": 0.23772727272727273
      },
      "relative": {
        "hit_at_10_percent": 0.0,
        "ndcg_at_10_percent": 0.0,
        "fresh_hit_at_10_percent": 0.0,
        "head_share_at_10_percent": 0.0
      },
      "stages": {
        "finite_scores": 360,
        "ledger": [
          {
            "candidate": "category-tower-80",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.02727272727272727,
              "ndcg_at_10": 0.012323290430588362,
              "fresh_hit_at_10": 0.03508771929824561,
              "head_share_at_10": 0.07181818181818182
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          },
          {
            "candidate": "category-tower-40",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.01818181818181818,
              "ndcg_at_10": 0.005855304256106641,
              "fresh_hit_at_10": 0.03508771929824561,
              "head_share_at_10": 0.056818181818181816
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          },
          {
            "candidate": "category-tower-20",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.04090909090909091,
              "ndcg_at_10": 0.019699458703707498,
              "fresh_hit_at_10": 0.0,
              "head_share_at_10": 0.03636363636363636
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          }
        ],
        "incumbent": "baseline",
        "reference": {
          "hit_at_10": 0.15,
          "ndcg_at_10": 0.07426696311942374,
          "fresh_hit_at_10": 0.07017543859649122,
          "head_share_at_10": 0.25136363636363634
        },
        "online_authorized": false
      },
      "paper_results": {
        "completed_offline_evaluations": 1586,
        "canonical_ledger_records": 3289,
        "launch_reviews": 9,
        "content_time_lift_sum_percent": 10.83
      },
      "scope": "CPU 小型机制验证：MovieLens 邻接关系不等于商品互补性；AutoLR 仅执行预定义候选的离线控制，不包含 LLM 调研/代码生成或线上发布。",
      "manifest_ref": "reproduction:autolr"
    },
    {
      "paper": {
        "arxiv_id": "2609.04871",
        "title": "AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems",
        "url": "https://arxiv.org/abs/2609.04871",
        "organization": "NetEase, Inc."
      },
      "dataset": {
        "name": "MovieLens 100K",
        "users": 220,
        "items": 360
      },
      "setup": {
        "adapter": "autolr",
        "same_split_and_candidates": true,
        "seed": 43
      },
      "baseline": {
        "name": "transition + content + popularity",
        "hit_at_10": 0.10909090909090909,
        "ndcg_at_10": 0.05400546778896325,
        "fresh_hit_at_10": 0.031746031746031744,
        "head_share_at_10": 0.23772727272727273
      },
      "method": {
        "name": "autolr executable offline mechanism",
        "hit_at_10": 0.10909090909090909,
        "ndcg_at_10": 0.05400546778896325,
        "fresh_hit_at_10": 0.031746031746031744,
        "head_share_at_10": 0.23772727272727273
      },
      "relative": {
        "hit_at_10_percent": 0.0,
        "ndcg_at_10_percent": 0.0,
        "fresh_hit_at_10_percent": 0.0,
        "head_share_at_10_percent": 0.0
      },
      "stages": {
        "finite_scores": 360,
        "ledger": [
          {
            "candidate": "category-tower-80",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.013636363636363636,
              "ndcg_at_10": 0.005855304256106641,
              "fresh_hit_at_10": 0.05263157894736842,
              "head_share_at_10": 0.041363636363636366
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          },
          {
            "candidate": "category-tower-40",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.013636363636363636,
              "ndcg_at_10": 0.006060215625416945,
              "fresh_hit_at_10": 0.03508771929824561,
              "head_share_at_10": 0.05545454545454546
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          },
          {
            "candidate": "category-tower-20",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.01818181818181818,
              "ndcg_at_10": 0.007266483708645933,
              "fresh_hit_at_10": 0.05263157894736842,
              "head_share_at_10": 0.056363636363636366
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          }
        ],
        "incumbent": "baseline",
        "reference": {
          "hit_at_10": 0.15,
          "ndcg_at_10": 0.07426696311942374,
          "fresh_hit_at_10": 0.07017543859649122,
          "head_share_at_10": 0.25136363636363634
        },
        "online_authorized": false
      },
      "paper_results": {
        "completed_offline_evaluations": 1586,
        "canonical_ledger_records": 3289,
        "launch_reviews": 9,
        "content_time_lift_sum_percent": 10.83
      },
      "scope": "CPU 小型机制验证：MovieLens 邻接关系不等于商品互补性；AutoLR 仅执行预定义候选的离线控制，不包含 LLM 调研/代码生成或线上发布。",
      "manifest_ref": "reproduction:autolr"
    },
    {
      "paper": {
        "arxiv_id": "2609.04871",
        "title": "AutoLR: Automating the Path from Research to Launch Review in Industrial Recommender Systems",
        "url": "https://arxiv.org/abs/2609.04871",
        "organization": "NetEase, Inc."
      },
      "dataset": {
        "name": "MovieLens 100K",
        "users": 220,
        "items": 360
      },
      "setup": {
        "adapter": "autolr",
        "same_split_and_candidates": true,
        "seed": 44
      },
      "baseline": {
        "name": "transition + content + popularity",
        "hit_at_10": 0.10909090909090909,
        "ndcg_at_10": 0.05400546778896325,
        "fresh_hit_at_10": 0.031746031746031744,
        "head_share_at_10": 0.23772727272727273
      },
      "method": {
        "name": "autolr executable offline mechanism",
        "hit_at_10": 0.10909090909090909,
        "ndcg_at_10": 0.05400546778896325,
        "fresh_hit_at_10": 0.031746031746031744,
        "head_share_at_10": 0.23772727272727273
      },
      "relative": {
        "hit_at_10_percent": 0.0,
        "ndcg_at_10_percent": 0.0,
        "fresh_hit_at_10_percent": 0.0,
        "head_share_at_10_percent": 0.0
      },
      "stages": {
        "finite_scores": 360,
        "ledger": [
          {
            "candidate": "category-tower-80",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.022727272727272728,
              "ndcg_at_10": 0.00932054141003343,
              "fresh_hit_at_10": 0.05263157894736842,
              "head_share_at_10": 0.1159090909090909
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          },
          {
            "candidate": "category-tower-40",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.013636363636363636,
              "ndcg_at_10": 0.005101809816596459,
              "fresh_hit_at_10": 0.017543859649122806,
              "head_share_at_10": 0.09636363636363636
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          },
          {
            "candidate": "category-tower-20",
            "events": [
              "selected",
              "verified",
              "offline-evaluated",
              "rejected"
            ],
            "metrics": {
              "hit_at_10": 0.02727272727272727,
              "ndcg_at_10": 0.012946370119032201,
              "fresh_hit_at_10": 0.03508771929824561,
              "head_share_at_10": 0.10136363636363636
            },
            "verdict": "rejected",
            "evidence": "arxiv:2609.05063; bounded local candidate, not live research",
            "reviews": [
              {
                "role": "budget",
                "feasible": true,
                "score": 1.0
              },
              {
                "role": "data-contract",
                "feasible": true,
                "score": 0.5
              }
            ],
            "online_authorized": false
          }
        ],
        "incumbent": "baseline",
        "reference": {
          "hit_at_10": 0.15,
          "ndcg_at_10": 0.07426696311942374,
          "fresh_hit_at_10": 0.07017543859649122,
          "head_share_at_10": 0.25136363636363634
        },
        "online_authorized": false
      },
      "paper_results": {
        "completed_offline_evaluations": 1586,
        "canonical_ledger_records": 3289,
        "launch_reviews": 9,
        "content_time_lift_sum_percent": 10.83
      },
      "scope": "CPU 小型机制验证：MovieLens 邻接关系不等于商品互补性；AutoLR 仅执行预定义候选的离线控制，不包含 LLM 调研/代码生成或线上发布。",
      "manifest_ref": "reproduction:autolr"
    }
  ],
  "aggregate_metrics": {
    "hit_at_10_mean": 0.10909090909090909,
    "hit_at_10_std": 0.0,
    "ndcg_at_10_mean": 0.05400546778896325,
    "ndcg_at_10_std": 0.0,
    "fresh_hit_at_10_mean": 0.031746031746031744,
    "fresh_hit_at_10_std": 0.0,
    "head_share_at_10_mean": 0.2377272727272727,
    "head_share_at_10_std": 0.0
  },
  "manifest_ref": "reproduction:autolr",
  "evaluation_protocol": {
    "tier": "l1_mechanism",
    "seeds": [
      42,
      43,
      44
    ],
    "formal_comparison": false,
    "claim_policy": "CPU mechanism diagnostic; no production or LLM capability claim"
  },
  "provenance": {
    "commit": "c943542b7685e4f24af8e4fce3a3b0f6f5df4b2b",
    "artifact_path": "docs/reproductions/2609.04871-autolr/metrics/movielens-100k-seeds42-44.json",
    "dataset_fingerprint": "b9913d93c70ffb6e98ee4e42085e312272b749fd4322d9c66eba02335eff8209",
    "command": "PYTHONPATH=src python scripts/regenerate_sep7_fidelity.py"
  },
  "correction": "Previous heuristic/oracle-based result withdrawn; this artifact executes corrected paths."
}