{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2605.30329", "kind": "paper", "url": "https://arxiv.org/abs/2605.30329"}, {"id": "gh:hosytuyen/soundnessbench", "kind": "repo", "url": "https://github.com/hosytuyen/SoundnessBench"}, {"kind": "website", "url": "https://hosytuyen.github.io/projects/SoundnessBench"}], "authors": ["Sy-Tuyen Ho", "Minghui Liu", "Huy Nghiem", "Hieu Tran", "Aakriti Agrawal", "Yongkyun Lee", "Mingyu Kang", "Andrew Adams Schoen", "Furong Huang"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2605.30329", "source_url": "https://arxiv.org/abs/2605.30329"}, "categories": ["AI Scientist", "General AI", "Language & Knowledge"], "description": {"en": "SoundnessBench evaluates LLMs' ability to judge the methodological soundness of research proposals. It contains 1,099 machine-learning research proposals reconstructed from ICLR submissions, labeled with reviewer soundness sub-scores and audited against source papers."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2605.30329", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2605.30329", "title": "SoundnessBench"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["University of Maryland, College Park"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2605.30329", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2605.30329", "source_url": "https://openalex.org/W7162769041"}, "key": "claire-radar:2605.30329", "modality": null, "name": "SoundnessBench", "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_soundnessbench_20ec76aa", "origin_source": "arxiv", "recognition_confidence": "0.95", "record_sha256": "0e198ed8afa1ba9e6fc29f83b68b66f64a57fefd2e4736c4837b8b2ec6a8fe81", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2605.30329"}, "publisher": null, "released": "2026-05-28", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2605.30329", "source_url": "https://arxiv.org/abs/2605.30329"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2605-30329", "source": "claire_radar", "source_benchmark_id": "2605.30329", "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": "https://github.com/hosytuyen/SoundnessBench", "githubScope": "benchmark_repo", "githubStars": 3, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 6, "hfPaperUrl": "https://huggingface.co/papers/2605.30329", "observedAt": "2026-10-04T15:31:48.030916Z", "signalStatus": {"githubStars": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}, "hfDatasetDownloads": {"state": "not_applicable"}, "hfDatasetLikes": {"state": "not_applicable"}, "hfPaperUpvotes": {"checkedAt": "2026-10-04T15:31:48.030916Z", "lastSuccessfulDate": "2026-09-30", "state": "stale"}}, "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": "4db51bc5b4fcf24b5daf1f72fcf0099f6fb6c16d65f28d8678cfc30898fb70da", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "SoundnessBench evaluates LLMs' ability to judge the methodological soundness of research proposals. It contains 1,099 machine-learning research proposals reconstructed from ICLR submissions, labeled with reviewer soundness sub-scores and audited against source papers.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["coined title prefix ending in Bench or Benchmark", "benchmark release stated in one sentence", "public artifact URL"], "snippet": "We introduce SoundnessBench, a curated benchmark of 1,099 machine-learning research proposals reconstructed from ICLR submissions, labeled with reviewer soundness sub-scores, and audited against source papers."}, "familyId": "bmf_479eac59a5d6", "firstSeenAt": "2026-08-19", "githubIndex": {"candidateCount": 0, "indexedAt": "2026-09-17T04:40:34.111665+00:00", "repositories": [{"evidence": [{"basis": "existing-code-link", "linkedUrl": "https://github.com/hosytuyen/SoundnessBench", "url": "https://github.com/hosytuyen/SoundnessBench"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/hosytuyen/SoundnessBench", "url": "https://huggingface.co/papers/2605.30329"}, {"basis": "linked-from-project", "linkedUrl": "https://github.com/hosytuyen/SoundnessBench", "url": "https://hosytuyen.github.io/projects/SoundnessBench"}], "scope": "benchmark_repo", "url": "https://github.com/hosytuyen/SoundnessBench", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_soundnessbench_20ec76aa", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/hosytuyen/SoundnessBench", "data": null, "hfPaper": "https://huggingface.co/papers/2605.30329", "paper": "https://arxiv.org/abs/2605.30329", "pdf": "https://arxiv.org/pdf/2605.30329", "project": "https://hosytuyen.github.io/projects/SoundnessBench", "report": "https://arxiv.org/abs/2605.30329"}, "metrics": [], "motivation": "Autonomous AI research agents aim to accelerate scientific discovery by automating the research pipeline, from hypothesis generation to peer review.", "name": "SoundnessBench", "oneLine": "SoundnessBench evaluates LLMs' ability to judge the methodological soundness of research proposals. It contains 1,099 machine-learning research proposals reconstructed from ICLR submissions, labeled with reviewer soundness sub-scores and audited against source papers.", "paperTitle": "SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2605.30329", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "hosytuyen", "organizationType": "academic-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/hosytuyen/SoundnessBench"}], "ranking": {}, "readiness": "Runnable", "recognitionConfidence": 0.95, "relation": "introduces", "releasedAt": "2026-05-28", "source": {"authors": ["Sy-Tuyen Ho", "Minghui Liu", "Huy Nghiem", "Furong Huang"], "categories": ["cs.LG"], "id": "2605.30329", "title": "SoundnessBench: Can Your AI Scientist Really Tell Good Research Ideas from Bad Ones?", "type": "arxiv", "url": "https://arxiv.org/abs/2605.30329"}, "sourceUpdatedAt": "2026-05-28T17:57:37Z", "topics": ["AI Scientist"], "venueAttempts": [], "whyItMatters": "Autonomous AI research agents need to evaluate research ideas before committing resources, but existing benchmarks do not test this bottleneck. SoundnessBench provides a standardized test for this capability."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
