{
  "paper": {
    "arxiv_id": "2608.00750",
    "title": "HRPO",
    "url": "https://arxiv.org/abs/2608.00750",
    "organization": "City University of Hong Kong / Kuaishou"
  },
  "dataset": {
    "name": "MovieLens-1M",
    "users": 280,
    "items": 440
  },
  "setup": {
    "seed": 42
  },
  "variants": {
    "sid sft": {
      "hit_at_10": 0.125,
      "ndcg_at_10": 0.06457068803903039,
      "head_share_at_10": 0.12964285714285714
    },
    "hrpo residual-return policy": {
      "hit_at_10": 0.1357142857142857,
      "ndcg_at_10": 0.06690137810958466,
      "head_share_at_10": 0.12928571428571428
    }
  },
  "relative": {
    "hit_at_10_percent": 8.571428571428562,
    "ndcg_at_10_percent": 3.6095171684487215,
    "head_share_at_10_percent": -0.27548209366391513
  },
  "mechanism": {
    "sid_depth": 9,
    "prefix_groups": 18,
    "residual_credit_std": 0.19689226393011655
  },
  "paper_results": {
    "short_drama_target_cost_percent": 0.168,
    "mini_game_target_cost_percent": 0.186,
    "fiction_target_cost_percent": 3.49
  },
  "scope": "执行层级 SID、prefix reward smoothing、residual credit 与 credit-to-go；未复刻快手广告策略网络和在线反馈。",
  "schema_version": 2,
  "manifest_ref": "reproduction:hrpo",
  "evaluation_protocol": {
    "tier": "l2_public_dataset",
    "seeds": [
      42
    ],
    "formal_comparison": false,
    "claim_policy": "single/few-seed smoke result; do not claim a stable improvement"
  },
  "provenance": {
    "artifact_path": "docs/reproductions/2608.00750-hrpo/metrics/public-seed42.json",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded",
    "dataset_fingerprint": "not recorded in historical artifact"
  }
}