{"record": {"aliases": [], "artifacts": [{"id": "arxiv:2607.25886", "kind": "paper", "url": "https://arxiv.org/abs/2607.25886"}, {"id": "gh:evolvent-ai/rsibench-data", "kind": "repo", "url": "https://github.com/evolvent-ai/RSIBench-Data"}], "authors": ["Fanqing Meng", "Lingxiao Du", "Qiguang Chen", "Ziqi Zhao", "Haocheng Lu", "Mengkang Hu", "Michael Qizhe Shieh"], "authors_reference": {"basis": "arxiv_metadata", "retrieved_at": "2026-11-03", "source_key": "claire-radar:2607.25886", "source_url": "https://arxiv.org/abs/2607.25886"}, "categories": ["Code & Software", "General AI", "Self-Evolution"], "description": {"en": "RSIBench-Data evaluates LLM agents as data-centric researchers, where agents iteratively revise training-data strategies for a fixed target model on six benchmarks, with real training and evaluation runs."}, "documents": [{"document_type": "registry_page", "id": "claire_radar:https://arxiv.org/abs/2607.25886", "source": "claire_radar", "source_url": "https://arxiv.org/abs/2607.25886", "title": "RSIBench-Data"}], "evidence_summary": {"document_count": 1, "model_count": null, "model_count_basis": null}, "institutions": ["National University of Singapore"], "institutions_reference": {"basis": "openalex_linked_institutions", "doi": "10.48550/arxiv.2607.25886", "retrieved_at": "2026-11-04", "source_key": "claire-radar:2607.25886", "source_url": "https://openalex.org/W7171709064"}, "key": "claire-radar:2607.25886", "modality": null, "name": "RSIBench-Data", "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_rsibench-data_1ace1cc0", "origin_source": "arxiv", "recognition_confidence": "1.0", "record_sha256": "cc997459987fead00654c84b0d3e1ce3b449c26aa2d38f81529cc8b964721549", "relation": "introduces", "review_state": "unreviewed", "source_url": "https://arxiv.org/abs/2607.25886"}, "publisher": null, "released": "2026-07-28", "released_reference": {"basis": "benchmark_release", "source_key": "claire-radar:2607.25886", "source_url": "https://arxiv.org/abs/2607.25886"}, "schema_version": 1, "sizes": [], "slug": "claire-radar-2607-25886", "source": "claire_radar", "source_benchmark_id": "2607.25886", "source_metadata": {"claire_radar": {"adoption30d": null, "aliases": [], "annotation": "Unknown", "applicationDomains": ["General AI"], "area": "Code & Software", "attention": {"asOf": "2026-10-05", "attemptedAt": "2026-10-04T15:31:48.030916Z", "githubRepo": "https://github.com/evolvent-ai/RSIBench-Data", "githubScope": "benchmark_repo", "githubStars": 168, "hfDailySubmittedAt": null, "hfDataset": null, "hfDatasetDownloads": null, "hfDatasetLikes": null, "hfPaperUpvotes": 0, "hfPaperUrl": "https://huggingface.co/papers/2607.25886", "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": {"observedAt": "2026-10-04T15:31:48.030916Z", "state": "fresh"}}, "status": "fresh"}, "capabilities": [], "confidence": "High", "construction": "Unknown", "constructionDetail": "Unknown — the indexer does not infer construction details without explicit source evidence.", "copyGeneration": {"generatedAt": "2026-08-20T15:22:10.044731Z", "inputHash": "8433842326a170ba278916ef9854aff1083d43680966b7a32be6b16c16d1ee6c", "model": "deepseek-v4-flash"}, "dataStatus": "primary-source-indexed", "demo": false, "description": "RSIBench-Data evaluates LLM agents as data-centric researchers, where agents iteratively revise training-data strategies for a fixed target model on six benchmarks, with real training and evaluation runs.", "domainCuration": {"method": "rules-v1", "state": "auto"}, "evidence": {"reasonCodes": ["exact coined title identity tied to benchmark evidence", "benchmark release stated in one sentence", "public artifact URL"], "snippet": "We introduce RSIBench-Data, a controlled benchmark of LLM agents as data-centric researchers with a fixed post-training stack."}, "familyId": "bmf_e3268eaf0064", "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/evolvent-ai/RSIBench-Data", "url": "https://github.com/evolvent-ai/RSIBench-Data"}, {"basis": "existing-metric-association", "linkedUrl": "https://github.com/evolvent-ai/RSIBench-Data", "url": "https://huggingface.co/papers/2607.25886"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/evolvent-ai/RSIBench-Data", "url": "https://arxiv.org/abs/2607.25886"}, {"basis": "linked-from-report", "linkedUrl": "https://github.com/evolvent-ai/RSIBench-Data.", "url": "https://arxiv.org/abs/2607.25886"}], "scope": "benchmark_repo", "url": "https://github.com/evolvent-ai/RSIBench-Data", "verification": "source-linked"}], "status": "linked"}, "heat": null, "id": "bm_rsibench-data_1ace1cc0", "indexedAt": "2026-08-19T11:05:13.395391Z", "industrySectors": [], "links": {"code": "https://github.com/evolvent-ai/RSIBench-Data", "data": null, "hfPaper": "https://huggingface.co/papers/2607.25886", "paper": "https://arxiv.org/abs/2607.25886", "pdf": "https://arxiv.org/pdf/2607.25886", "project": null, "report": "https://arxiv.org/abs/2607.25886"}, "metrics": [], "motivation": "Recursive self-improvement requires turning evidence of model failures into better models.", "name": "RSIBench-Data", "oneLine": "RSIBench-Data evaluates LLM agents as data-centric researchers, where agents iteratively revise training-data strategies for a fixed target model on six benchmarks, with real training and evaluation runs.", "paperTitle": "RSIBench-Data: Benchmarking Data-Centric Research for Recursive Self-Improvement", "primaryDomain": "General AI", "publication": {"evidence": null, "evidenceLevel": "unverified", "evidenceUrl": "https://arxiv.org/abs/2607.25886", "source": "arxiv-metadata", "status": "unverified", "venue": null, "verifiedAt": "2026-10-14T01:46:58.058199Z"}, "publications": [], "publishers": [{"name": "evolvent-ai", "organizationType": "company-research-lab", "role": "benchmark-publisher", "sourceUrl": "https://github.com/evolvent-ai/RSIBench-Data"}], "ranking": {"90d": {"confidence": "Medium", "coverage": 0.7, "growth": {"components": {"githubStars": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfDatasetDownloads": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}, "hfPaperUpvotes": {"normalization": "percentile", "normalized": null, "percentile": null, "value": null}}, "confidence": "Low", "coverage": 0.0, "elapsedDays": null, "historyDate": null, "rank": null, "score": null, "windowDays": 90}, "level": {"components": {"githubStars": {"normalization": "log-max", "normalized": 0.6847492627616909, "value": 168}, "hfDatasetDownloads": {"normalization": "log-max", "normalized": null, "value": null}, "hfPaperUpvotes": {"normalization": "log-max", "normalized": 0.0, "value": 0}}, "confidence": "Medium", "coverage": 0.7, "rank": 43, "score": 53}, "method": "attention-ranking-v13", "rank": 43, "score": 53}}, "readiness": "Runnable", "recognitionConfidence": 1.0, "relation": "introduces", "releasedAt": "2026-07-28", "source": {"authors": ["Fanqing Meng", "Lingxiao Du", "Qiguang Chen", "Ziqi Zhao", "Haocheng Lu", "Mengkang Hu", "Michael Qizhe Shieh"], "categories": ["cs.SE", "cs.CL"], "id": "2607.25886", "title": "RSIBench-Data: Benchmarking Data-Centric Research for Recursive Self-Improvement", "type": "arxiv", "url": "https://arxiv.org/abs/2607.25886"}, "sourceUpdatedAt": "2026-07-28T15:46:41Z", "topics": ["Self-Evolution"], "venueAttempts": [], "whyItMatters": "It isolates research capability from engineering, showing that current agents can improve from feedback but inconsistently. This provides an auditable testbed for capabilities needed in recursive self-improvement."}}}, "schema_version": 1, "scores_by_source": {}, "siblings": []}
