{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.24043", "kind": "paper", "url": "https://arxiv.org/abs/2605.24043"}, {"id": "gh:scientific-discovery/llm-autoscilab", "kind": "repo", "url": "https://github.com/scientific-discovery/LLM-AutoSciLab"}], "categories": ["AI Scientist", "General AI", "Language & Knowledge"], "description": {"en": "ActiveSciBench evaluates budget-constrained, closed-loop scientific discovery across enzyme-kinetics and gene-regulatory-network tasks."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.24043", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.24043", "title": "ActiveSciBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "key": "claire-radar:2605.24043", "modality": null, "name": "ActiveSciBench", "openness": {"code_license": null, "data_license": null, "evidence": [], "status": "unknown"}, "provenance": {"admission_policy_version": "2026-09-09.1", "confidence": "Medium", "crawl_bundle": "claire_radar_2026-10-14", "crawled_at": "2026-09-25T04:40:22Z", "data_status": "primary-source-indexed", "display_eligible": "true", "origin_record_id": "bm_activescibench_04dc839f", "origin_source": "arxiv", "recognition_confidence": "0.55", "record_sha256": "a4b4e3ce04f68d2138cb62cfc503b13411da00d35130c08a472156c759e729cc", "relation": "evaluates_only", "review_model": "codex-cli-default", "review_state": "ai-reviewed", "reviewed_at": "2026-09-24T19:39:58.745325Z", "source_url": "https://arxiv.org/abs/2605.24043"}, "publisher": null, "released": "2026-05-21", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.24043", "source_url": "https://arxiv.org/abs/2605.24043"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-24043", "source": "claire_radar", "source_benchmark_id": "2605.24043", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Language & Knowledge", "attentionForecast": {"confidence": "Medium", "horizon": "7d", "reason": "The benchmark addresses active scientific discovery with a public implementation and two domain-specific task collections, which may attract near-term research interest.", "score": 7}, "capabilities": [], "confidence": "Medium", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-09-24T19:35:17.646109Z", "inputHash": "0cb89ee7a02500bfa94a3280afe0924b2c4748af80a859895ced23c0029c89b7", "model": "codex-cli-default"}, "curation": {"admissionPolicyVersion": "2026-09-09.1", "canonicalNameEvidence": "we introduce ActiveSciBench", "canonicalNameSource": "abstract", "decisionReason": "The source formally names the benchmark, defines repeatable task collections and comparable accuracy and graph-recovery measures, and provides an explicitly available public code-and-data path.", "evaluationEvidence": {"reuse": "Code and data are available at: https://github.com/scientific-discovery/LLM-AutoSciLab", "scoring": "achieving 67.6% and 35.1% symbolic accuracy on NewtonBench and ActiveSciBench-Chem, respectively, and 31.1% exact graph recovery on ActiveSciBench-GRN.", "task": "These datasets model discovery as a budget-constrained process requiring adaptive experiment design, variable selection, and recovery of true mechanisms."}, "model": "codex-cli-default", "reviewedAt": "2026-09-24T19:39:58.745325Z", "state": "ai-reviewed"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "ActiveSciBench evaluates budget-constrained, closed-loop scientific discovery across enzyme-kinetics and gene-regulatory-network tasks.", "displayEligible": true, "domainCuration": {"method": "rules-v1", "state": "auto"}, "evaluationMode": "public_reusable", "evidence": {"reasonCodes": ["exact named benchmark artifact released in abstract", "evaluation protocol evidence", "public artifact URL", "named entity is a model, method, framework, or dataset used in evaluation"], "snippet": "To evaluate dynamic, closed-loop scientific discovery with active data acquisition, we introduce ActiveSciBench, comprising two datasets: ActiveSciBench-Chem with 57 enzyme-kinetics tasks and ActiveSciBench-GRN with 45 gene-regulatory-network tasks."}, "familyId": "bmf_8c1f7bd85643", "firstSeenAt": "2026-08-19", "heat": null, "id": "bm_activescibench_04dc839f", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/scientific-discovery/LLM-AutoSciLab", "data": null, "paper": "https://arxiv.org/abs/2605.24043", "pdf": "https://arxiv.org/pdf/2605.24043", "project": null, "report": "https://arxiv.org/abs/2605.24043"}, "metrics": [], "motivation": "Scientific discovery is a closed-loop process in which hypotheses guide data acquisition and observations refine the hypothesis space.", "name": "ActiveSciBench", "oneLine": "ActiveSciBench evaluates budget-constrained, closed-loop scientific discovery across enzyme-kinetics and gene-regulatory-network tasks.", "paperTitle": "LLM-AutoSciLab: Closed-Loop Scientific Discovery via Active Experimentation with LLMs", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.24043", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-08-19T11:05:13.395391Z"}, "publications": [], "publishers": [{"name": "scientific-discovery", "organizationType": "community", "role": "benchmark-publisher", "sourceUrl": "https://github.com/scientific-discovery/LLM-AutoSciLab"}], "readiness": "Runnable", "recognitionConfidence": 0.55, "relation": "evaluates_only", "releasedAt": "2026-05-21", "reviewLifecycle": {"state": "pending"}, "source": {"authors": ["Sanchit Kabra", "Nikhil Abhyankar", "Saaketh Desai", "Prasad Iyer", "Chandan K Reddy"], "categories": ["cs.LG", "cs.AI"], "id": "2605.24043", "title": "LLM-AutoSciLab: Closed-Loop Scientific Discovery via Active Experimentation with LLMs", "type": "arxiv", "url": "https://arxiv.org/abs/2605.24043"}, "sourceUpdatedAt": "2026-05-21T20:30:56Z", "topics": ["AI Scientist"], "venueAttempts": [], "whyItMatters": "It tests whether systems can adaptively select experiments and recover underlying mechanisms rather than infer only from fixed observations."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
