{
  "paper": "2605.27856",
  "adapter": "pinterest-ads-llm",
  "dataset": {
    "name": "MiniOneRec Office",
    "train_rows": 12000,
    "advertisers": 48,
    "validation_users": 32,
    "test_users": 32
  },
  "setup": {
    "model": "HuggingFaceTB/SmolLM2-135M-Instruct",
    "seed": 42,
    "sft_steps": 48,
    "grpo_steps": 16,
    "tower_steps": 160,
    "selected": "sft"
  },
  "test": {
    "sft": {
      "recall_at_1": 0.0625,
      "recall_at_5": 0.0625,
      "recall_at_20": 0.125,
      "mean_rank": 36.5625
    },
    "sft_grpo": {
      "recall_at_1": 0.0625,
      "recall_at_5": 0.0625,
      "recall_at_20": 0.125,
      "mean_rank": 34.625
    }
  },
  "downstream": {
    "retrieval_recall_at_50_baseline": 0.125,
    "retrieval_recall_at_50_complementary": 0.125,
    "selected_llm_quota": 0,
    "ranking_auc_baseline": 0.6261307566,
    "ranking_auc_with_llm_feature": 0.6423211349,
    "ranking_auc_relative_percent": 2.5857823018
  },
  "conclusion": "Validation rejects GRPO and any positive complementary-retrieval quota; the LLM advertiser score gives a small sampled-ranking AUC gain.",
  "schema_version": 2,
  "manifest_ref": "reproduction:pinterest-ads-llm",
  "evaluation_protocol": {
    "tier": "l3_paper_pipeline",
    "seeds": [
      42
    ],
    "formal_comparison": false,
    "claim_policy": "single/few-seed smoke result; do not claim a stable improvement"
  },
  "provenance": {
    "artifact_path": "docs/reproductions/2605.27856-pinterest-ads-llm/metrics/office-seed42.json",
    "historical_migration": "historical-metrics-v2-2026-08-09",
    "original_code_commit": "not recorded",
    "dataset_fingerprint": "not recorded in historical artifact"
  }
}