{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.24117", "kind": "paper", "url": "https://arxiv.org/abs/2605.24117"}], "authors": ["Yingtie Lei", "Zhongwei Wan", "Jiankun Zhang", "Samiul Alam", "Zixuan Zhong", "Peizhou Huang", "Xin Wang", "Jingxuan Zhang", "Donghao Zhou", "Yunta Hsieh", "Zhihao Dou", "Hui Shen", "Yan Xu", "Dimitrios Dimitriadis", "Tuo Zhang", "Mi Zhang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.24117", "source_url": "https://arxiv.org/abs/2605.24117"}, "categories": ["General AI", "Language & Knowledge", "cs.AI"], "description": {"en": "SkillEvolBench evaluates whether LLM agents can distill episodic experience into reusable procedural skills. It includes 180 tasks across six environments, with acquisition tasks and frozen deployment tasks testing context shift, adversarial shortcuts, and composition. Scoring uses success rates and other agent metrics."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.24117", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.24117", "title": "SkillEvolBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["The Ohio State University", "University of Chicago", "University College London", "University of Michigan", "Chinese University of Hong Kong", "Case Western Reserve University"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2605.24117", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2605.24117", "source_url": "https://openalex.org/W7162468227"}, "key": "claire-radar:2605.24117", "modality": null, "name": "SkillEvolBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"confidence": "High", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-10-14T09:44:51Z", "data_status": "primary-source-indexed", "display_eligible": "unknown", "origin_record_id": "bm_skillevolbench_997ef790", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "de13df99bf427e59d33b083af6cdce19a7af789cfc96d745bb8fd2ffc954cd2d", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.24117"}, "publisher": null, "released": "2026-05-22", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.24117", "source_url": "https://arxiv.org/abs/2605.24117"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-24117", "source": "claire_radar", "source_benchmark_id": "2605.24117", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": null, "githubScope": null, "githubStars": null, "hfDailySubmittedAt": "2026-05-26T00:00:00.000Z", "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 19, "hfPaperUrl": "https://huggingface.co/papers/2605.24117", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"state": "not_applicable"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Transform Existing", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "67b22643bad860b49d7efe0353d87e1c99be64e10066d727165156af7747dd3c", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "SkillEvolBench evaluates whether LLM agents can distill episodic experience into reusable procedural skills. It includes 180 tasks across six environments, with acquisition tasks and frozen deployment tasks testing context shift, adversarial shortcuts, and composition. Scoring uses success rates and other agent metrics.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "evaluation protocol evidence"], "snippet": "We introduce SkillEvolBench, a diagnostic benchmark for evaluating this step from experience reuse to skill formation."}, "familyId": "bmf_0d32e638a5b1", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [], "status": "not-found"}, "heat": null, "id": "bm_skillevolbench_997ef790", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": null, "data": null, "hfPaper": "https://huggingface.co/papers/2605.24117", "paper": "https://arxiv.org/abs/2605.24117", "pdf": "https://arxiv.org/pdf/2605.24117", "project": null, "report": "https://arxiv.org/abs/2605.24117"}, "metrics": [], "motivation": "Large language model (LLM) agents accumulate rich episodic trajectories while solving real-world tasks, but it remains unclear whether such experience can be distilled into reusable procedural skills.", "name": "SkillEvolBench", "oneLine": "SkillEvolBench evaluates whether LLM agents can distill episodic experience into reusable procedural skills. It includes 180 tasks across six environments, with acquisition tasks and frozen deployment tasks testing context shift, adversarial shortcuts, and composition. Scoring uses success rates and other agent metrics.", "paperTitle": "SkillEvolBench: Benchmarking the Evolution from Episodic Experience to Procedural Skills", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.24117", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "ranking": {}, "readiness": "Paper only", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-05-22", "source": {"authors": ["Yingtie Lei", "Zhongwei Wan", "Jiankun Zhang", "Samiul Alam", "Zixuan Zhong", "Peizhou Huang", "Xin Wang", "Jingxuan Zhang", "Donghao Zhou", "Yunta Hsieh", "Zhihao Dou", "Hui Shen", "Yan Xu", "Dimitrios Dimitriadis", "Tuo Zhang", "Mi Zhang"], "categories": ["cs.AI"], "id": "2605.24117", "title": "SkillEvolBench: Benchmarking the Evolution from Episodic Experience to Procedural Skills", "type": "arxiv", "url": "https://arxiv.org/abs/2605.24117"}, "sourceUpdatedAt": "2026-05-22T18:23:31Z", "topics": ["cs.AI"], "venueAttempts": [], "whyItMatters": "It is unclear whether LLM agents can form durable procedural skills from experience. SkillEvolBench provides a diagnostic testbed comparing skill-based learning against raw-trajectory reuse, informing agent design and training."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
