{ "paper": { "arxiv_id": "2601.18664", "title": "S2GR: Stepwise Semantic-Guided Reasoning in Latent Space for Generative Recommendation", "url": "https://arxiv.org/abs/2601.18664", "organization": "Kuaishou Technology" }, "dataset": { "name": "MovieLens 100K", "users": 220, "items": 360 }, "setup": { "adapter": "s2gr", "same_split_and_candidates": true }, "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": "s2gr core mechanism (validation blend=0.1)", "hit_at_10": 0.09545454545454546, "ndcg_at_10": 0.05462715341481358, "fresh_hit_at_10": 0.06349206349206349, "head_share_at_10": 0.17681818181818182 }, "relative": { "hit_at_10_percent": -12.499999999999991, "ndcg_at_10_percent": 1.1511531170874825, "fresh_hit_at_10_percent": 100.0, "head_share_at_10_percent": -25.62141491395794 }, "stages": { "finite_scores": 360, "score_std": 0.1232491885824412, "reasoning_steps": 4, "semantic_code_levels": 4 }, "paper_results": { "online_core_metric_lift_percent": 5.25 }, "scope": "执行论文可由公开 MovieLens 特征审计的核心计算；不复刻私有日志、生产模型或线上服务栈。", "manifest_ref": "reproduction:s2gr", "schema_version": 2, "manifest": { "adapter_key": "s2gr", "arxiv_id": "2601.18664", "title": "S2GR: Stepwise Semantic-Guided Reasoning in Latent Space for Generative Recommendation", "paper_url": "https://arxiv.org/abs/2601.18664", "track": "recommendation", "organization": "Kuaishou Technology", "published": "2026-01-26", "code_url": null, "topics": [ "generative-recommendation", "latent-reasoning", "semantic-id" ], "local_code_dir": "src/auto_research/reproductions/s2gr", "fidelity": "core_mechanism", "evaluation_tier": "l2_public_dataset", "datasets": [ "MovieLens 100K" ], "baseline": "transition + content + popularity", "metrics": [ "hit_at_10", "ndcg_at_10", "fresh_hit_at_10", "head_share_at_10" ], "default_seeds": [ 42, 43, 44 ], "budget": "220 users / 360 items; validation-only blend selection", "device_capabilities": [ "cpu" ], "requires_gpu_validation": false, "gpu_validation_artifact": null, "online_evidence": [ { "product": "Kuaishou recommendation", "metric": "Online core metric", "lift_percent": 5.25, "traffic": "production online A/B", "source_url": "https://arxiv.org/html/2601.18664v1", "source_location": "Section 5.4", "retrieved_at": "2026-09-05" } ], "selection_exception": null, "evolve_operators": [ "head:stepwise-semantic-reasoning" ] }, "provenance": { "created_at": "2026-09-05T08:50:48.734067+00:00", "code_commit": "2e0122e2076fde02d644c754315875a68f7d87c6", "python": "3.12.9", "platform": "macOS-26.5.2-arm64-arm-64bit", "dataset_dir": "/Users/bytedance/Documents/git_daiwk/auto-research/data", "dataset_fingerprint": "e66c261317a7b3179720ef3e3f5f79d7a204b135d1f217e2911b153ad2ce50a0", "packages": { "auto-research": "0.1.0", "numpy": "2.5.2", "torch": "2.13.0", "transformers": "5.14.1" }, "artifact_path": "docs/reproductions/2601.18664-s2gr/metrics/public-seeds42-44.json" }, "evaluation_protocol": { "tier": "l2_public_dataset", "tier_label": "L2 公开数据集训练", "seeds": [ 42, 43, 44 ], "budget": "paper-specific", "formal_comparison": true, "claim_policy": "formal multi-seed comparison" }, "seed_results": [ { "paper": { "arxiv_id": "2601.18664", "title": "S2GR: Stepwise Semantic-Guided Reasoning in Latent Space for Generative Recommendation", "url": "https://arxiv.org/abs/2601.18664", "organization": "Kuaishou Technology" }, "dataset": { "name": "MovieLens 100K", "users": 220, "items": 360 }, "setup": { "adapter": "s2gr", "same_split_and_candidates": true }, "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": "s2gr core mechanism (validation blend=0.1)", "hit_at_10": 0.09545454545454546, "ndcg_at_10": 0.05462715341481358, "fresh_hit_at_10": 0.06349206349206349, "head_share_at_10": 0.17681818181818182 }, "relative": { "hit_at_10_percent": -12.499999999999991, "ndcg_at_10_percent": 1.1511531170874825, "fresh_hit_at_10_percent": 100.0, "head_share_at_10_percent": -25.62141491395794 }, "stages": { "finite_scores": 360, "score_std": 0.1232491885824412, "reasoning_steps": 4, "semantic_code_levels": 4 }, "paper_results": { "online_core_metric_lift_percent": 5.25 }, "scope": "执行论文可由公开 MovieLens 特征审计的核心计算；不复刻私有日志、生产模型或线上服务栈。", "manifest_ref": "reproduction:s2gr", "seed": 42 }, { "paper": { "arxiv_id": "2601.18664", "title": "S2GR: Stepwise Semantic-Guided Reasoning in Latent Space for Generative Recommendation", "url": "https://arxiv.org/abs/2601.18664", "organization": "Kuaishou Technology" }, "dataset": { "name": "MovieLens 100K", "users": 220, "items": 360 }, "setup": { "adapter": "s2gr", "same_split_and_candidates": true }, "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": "s2gr core mechanism (validation blend=0.1)", "hit_at_10": 0.09545454545454546, "ndcg_at_10": 0.05462715341481358, "fresh_hit_at_10": 0.06349206349206349, "head_share_at_10": 0.17681818181818182 }, "relative": { "hit_at_10_percent": -12.499999999999991, "ndcg_at_10_percent": 1.1511531170874825, "fresh_hit_at_10_percent": 100.0, "head_share_at_10_percent": -25.62141491395794 }, "stages": { "finite_scores": 360, "score_std": 0.1232491885824412, "reasoning_steps": 4, "semantic_code_levels": 4 }, "paper_results": { "online_core_metric_lift_percent": 5.25 }, "scope": "执行论文可由公开 MovieLens 特征审计的核心计算；不复刻私有日志、生产模型或线上服务栈。", "manifest_ref": "reproduction:s2gr", "seed": 43 }, { "paper": { "arxiv_id": "2601.18664", "title": "S2GR: Stepwise Semantic-Guided Reasoning in Latent Space for Generative Recommendation", "url": "https://arxiv.org/abs/2601.18664", "organization": "Kuaishou Technology" }, "dataset": { "name": "MovieLens 100K", "users": 220, "items": 360 }, "setup": { "adapter": "s2gr", "same_split_and_candidates": true }, "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": "s2gr core mechanism (validation blend=0.1)", "hit_at_10": 0.09545454545454546, "ndcg_at_10": 0.05462715341481358, "fresh_hit_at_10": 0.06349206349206349, "head_share_at_10": 0.17681818181818182 }, "relative": { "hit_at_10_percent": -12.499999999999991, "ndcg_at_10_percent": 1.1511531170874825, "fresh_hit_at_10_percent": 100.0, "head_share_at_10_percent": -25.62141491395794 }, "stages": { "finite_scores": 360, "score_std": 0.1232491885824412, "reasoning_steps": 4, "semantic_code_levels": 4 }, "paper_results": { "online_core_metric_lift_percent": 5.25 }, "scope": "执行论文可由公开 MovieLens 特征审计的核心计算；不复刻私有日志、生产模型或线上服务栈。", "manifest_ref": "reproduction:s2gr", "seed": 44 } ], "aggregate_metrics": { "dataset.users": { "mean": 220.0, "std": 0.0, "ci95": 0.0, "n": 3 }, "dataset.items": { "mean": 360.0, "std": 0.0, "ci95": 0.0, "n": 3 }, "baseline.hit_at_10": { "mean": 0.10909090909090909, "std": 0.0, "ci95": 0.0, "n": 3 }, "baseline.ndcg_at_10": { "mean": 0.05400546778896325, "std": 0.0, "ci95": 0.0, "n": 3 }, "baseline.fresh_hit_at_10": { "mean": 0.031746031746031744, "std": 0.0, "ci95": 0.0, "n": 3 }, "baseline.head_share_at_10": { "mean": 0.2377272727272727, "std": 0.0, "ci95": 0.0, "n": 3 }, "method.hit_at_10": { "mean": 0.09545454545454546, "std": 0.0, "ci95": 0.0, "n": 3 }, "method.ndcg_at_10": { "mean": 0.05462715341481358, "std": 0.0, "ci95": 0.0, "n": 3 }, "method.fresh_hit_at_10": { "mean": 0.06349206349206349, "std": 0.0, "ci95": 0.0, "n": 3 }, "method.head_share_at_10": { "mean": 0.17681818181818185, "std": 0.0, "ci95": 0.0, "n": 3 }, "relative.hit_at_10_percent": { "mean": -12.499999999999991, "std": 0.0, "ci95": 0.0, "n": 3 }, "relative.ndcg_at_10_percent": { "mean": 1.1511531170874825, "std": 0.0, "ci95": 0.0, "n": 3 }, "relative.fresh_hit_at_10_percent": { "mean": 100.0, "std": 0.0, "ci95": 0.0, "n": 3 }, "relative.head_share_at_10_percent": { "mean": -25.62141491395794, "std": 0.0, "ci95": 0.0, "n": 3 }, "stages.finite_scores": { "mean": 360.0, "std": 0.0, "ci95": 0.0, "n": 3 }, "stages.score_std": { "mean": 0.1232491885824412, "std": 0.0, "ci95": 0.0, "n": 3 }, "stages.reasoning_steps": { "mean": 4.0, "std": 0.0, "ci95": 0.0, "n": 3 }, "stages.semantic_code_levels": { "mean": 4.0, "std": 0.0, "ci95": 0.0, "n": 3 }, "paper_results.online_core_metric_lift_percent": { "mean": 5.25, "std": 0.0, "ci95": 0.0, "n": 3 } } }