{"record": {"aliases": ["Beyond Factual Knowledge: Benchmarking and Learning Step-Level Procedural Rule Reasoning in Large Language Models"], "artifacts": [{"id": "arxiv:2608.22753", "kind": "paper", "url": "https://arxiv.org/abs/2608.22753v1"}, {"id": "gh:sharkspicy-nlp/beyond-factual-knowledge", "kind": "repo", "url": "https://github.com/SharkSpicy-NLP/Beyond-Factual-Knowledge"}], "authors": ["Bohan Yu", "Pengfei Cao", "Chen Han", "Chenxi Zhou", "Zhiheng Zhang", "Zhiyang Xie", "Wenhao Teng", "Xiangwen Liao", "Jun Zhao", "Kang Liu"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2608.22753", "source_url": "https://arxiv.org/abs/2608.22753"}, "categories": ["Factuality", "General AI", "Language & Knowledge", "Reasoning"], "description": {"en": "Evaluates step-level procedural rule reasoning with single-rule, parallel multi-rule, and multi-hop scenarios over large rule pools."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2608.22753v1", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2608.22753v1", "title": "RuleWorld"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2608.22753", "modality": null, "name": "RuleWorld", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_beyond-factual-knowledge-benchmarking-and-_bf5f8184", "origin_source": "arxiv", "recognition_confidence": "0.7", "record_sha256": "9409331adc8746aa935e011a9b34ec2db03fbfd77b4fd4e209ad28cd483960da", "relation": "introduces", "review_model": "deepseek-v4-pro", "review_state": "ai-reviewed", "reviewed_at": "2026-08-27T04:18:56.318436Z", "source_url": "https://arxiv.org/abs/2608.22753v1"}, "publisher": null, "released": "2026-08-24", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2608.22753", "source_url": "https://arxiv.org/abs/2608.22753v1"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2608-22753", "source": "claire_radar", "source_benchmark_id": "2608.22753", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": ["Beyond Factual Knowledge: Benchmarking and Learning Step-Level Procedural Rule Reasoning in Large Language Models"], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-08-26", "attemptedAt": "2026-08-25T16:48:55.714492Z", "githubRepo": "https://github.com/SharkSpicy-NLP/Beyond-Factual-Knowledge", "githubScope": "benchmark_repo", "githubStars": 4, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2608.22753", "observedAt": "2026-08-25T16:48:55.714492Z", "signalStatus": {"githubStars": {"observedAt": "2026-08-25T16:48:55.714492Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-08-25T16:48:55.714492Z", "state": "fresh"}}, "status": "fresh"}, "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "Large-scale dataset and a new training framework with strong reported gains are likely to attract attention in the NLP community.", "score": 62}, "capabilities": ["Reasoning", "Factuality"], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-27T04:18:56.318436Z", "inputHash": "e9f000a22a56b1fbf9ece448403765db9dde6b1989c9278f53b621af035dd44d", "model": "deepseek-v4-pro"}, "curation": {"canonicalNameEvidence": "we introduce RuleWorld, a large-scale benchmark that reformulates rules as globally reusable abstract units", "canonicalNameSource": "abstract", "decisionReason": "Formally named benchmark with dataset released on Hugging Face and code on GitHub, providing a reusable evaluation protocol and scoring via QA accuracy.", "model": "deepseek-v4-pro", "reviewedAt": "2026-08-27T04:18:56.318436Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "Evaluates step-level procedural rule reasoning with single-rule, parallel multi-rule, and multi-hop scenarios over large rule pools.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["benchmark term in abstract", "benchmark release stated in one sentence", "evaluation protocol evidence", "public artifact URL"], "snippet": "To evaluate this capability, we introduce RuleWorld, a large-scale benchmark that reformulates rules as globally reusable abstract units rather than instance-specific facts."}, "familyId": "bmf_ace1cc273de1", "firstSeenAt": "2026-08-25", "heat": null, "id": "bm_beyond-factual-knowledge-benchmarking-and-_bf5f8184", "indexedAt": "2026-08-25T16:21:05.134052Z", "industrySectors": [], "links": {"code": "https://github.com/SharkSpicy-NLP/Beyond-Factual-Knowledge", "data": null, "paper": "http://arxiv.org/abs/2608.22753v1", "pdf": "https://arxiv.org/pdf/2608.22753v1", "project": null, "report": "http://arxiv.org/abs/2608.22753v1"}, "metrics": [], "motivation": "Large language models (LLMs) excel at text understanding and generation, yet still struggle to reliably understand and apply externally provided procedural rules at scale.", "name": "RuleWorld", "oneLine": "Evaluates step-level procedural rule reasoning with single-rule, parallel multi-rule, and multi-hop scenarios over large rule pools.", "paperTitle": "Beyond Factual Knowledge: Benchmarking and Learning Step-Level Procedural Rule Reasoning in Large Language Models", "primaryDomain": "General AI", "publication": {"evidence": "Accepted by EMNLP 2026 Findings", "evidenceLevel": "author-claim", "evidenceUrl": "http://arxiv.org/abs/2608.22753v1", "source": "arxiv-comments", "status": "acceptance_claimed", "venue": "EMNLP 2026 Findings", "verifiedAt": "2026-08-25T16:21:05.134052Z"}, "publications": [], "publishers": [{"name": "RuleWorld contributors", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/SharkSpicy-NLP/Beyond-Factual-Knowledge"}], "ranking": {"30d": {"confidence": "High", "coverage": 0.85, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 30}, "level": {"components": {"githubStars": {"percentile": 0.6766917293233082, "value": 4}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": 0.1717171717171717, "value": 0}}, "confidence": "High", "coverage": 0.85, "rank": 69, "score": 50}, "method": "attention-ranking-v11", "rank": 69, "score": 50}, "90d": {"confidence": "Medium", "coverage": 0.7, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"percentile": 0.4832869080779944, "value": 4}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": 0.2119047619047619, "value": 0}}, "confidence": "Medium", "coverage": 0.7, "rank": 229, "score": 45}, "method": "attention-ranking-v11", "rank": 229, "score": 45}, "today": {"confidence": "High", "coverage": 0.95, "growth": {"components": {"githubStars": {"percentile": null, "value": null}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": null, "value": null}, "llmAttentionForecast": {"percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": 2, "historyDate": "2026-08-24", "rank": null, "score": null, "windowDays": 1}, "level": {"components": {"githubStars": {"percentile": 0.5, "value": 4}, "hfDatasetDownloads": {"percentile": null, "value": null}, "hfPaperUpvotes": {"percentile": 0.5, "value": 0}, "llmAttentionForecast": {"percentile": 0.5, "value": 50}}, "confidence": "High", "coverage": 0.95, "rank": 7, "score": 50}, "method": "attention-ranking-v11", "rank": 7, "score": 50}}, "readiness": "Runnable", "recognitionConfidence": 0.7, "relation": "introduces", "releasedAt": "2026-08-24", "source": {"authors": ["Bohan Yu", "Pengfei Cao", "Chen Han", "Chenxi Zhou", "Zhiheng Zhang", "Zhiyang Xie", "Wenhao Teng", "Xiangwen Liao", "Jun Zhao", "Kang Liu"], "categories": ["cs.CL"], "id": "2608.22753", "title": "Beyond Factual Knowledge: Benchmarking and Learning Step-Level Procedural Rule Reasoning in Large Language Models", "type": "arxiv", "url": "http://arxiv.org/abs/2608.22753v1"}, "sourceUpdatedAt": "2026-08-24T03:22:45Z", "topics": ["Reasoning"], "venueAttempts": [{"decisionRaw": "Accepted by EMNLP 2026 Findings", "evidence": [{"level": "author-claim", "observedAt": "2026-08-25T16:21:05.134052Z", "rawValue": "Accepted by EMNLP 2026 Findings", "sourceType": "arxiv-comments", "sourceUrl": "http://arxiv.org/abs/2608.22753v1"}], "reviewStatus": "accepted", "venueName": "EMNLP 2026 Findings"}], "whyItMatters": "Tests whether models can apply externally provided rules at scale, a capability needed for reliable procedural reasoning in real-world tasks."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
