{"bundle_type":"pith_open_graph_bundle","bundle_version":"1.0","pith_number":"pith:2026:QXYMIEQDSE3D2RQXG2QGJED4FX","short_pith_number":"pith:QXYMIEQD","canonical_record":{"source":{"id":"2605.07022","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T23:08:18Z","cross_cats_sorted":[],"title_canon_sha256":"4b83343cffc37ad48fdb397209199bc51669e3276667c6f46ce5b357c140b019","abstract_canon_sha256":"534fe76e0dfbf7c293edda4666f9e3c40e246050a72cfa8d9e2bd17bae455c6d"},"schema_version":"1.0"},"canonical_sha256":"85f0c4120391363d461736a064907c2dfd47c8b5b15737f9d1667bc8ee7640bd","source":{"kind":"arxiv","id":"2605.07022","version":2},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.07022","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"arxiv_version","alias_value":"2605.07022v2","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.07022","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"pith_short_12","alias_value":"QXYMIEQDSE3D","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"pith_short_16","alias_value":"QXYMIEQDSE3D2RQX","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"pith_short_8","alias_value":"QXYMIEQD","created_at":"2026-05-20T00:02:12Z"}],"events":[{"event_type":"record_created","subject_pith_number":"pith:2026:QXYMIEQDSE3D2RQXG2QGJED4FX","target":"record","payload":{"canonical_record":{"source":{"id":"2605.07022","kind":"arxiv","version":2},"metadata":{"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T23:08:18Z","cross_cats_sorted":[],"title_canon_sha256":"4b83343cffc37ad48fdb397209199bc51669e3276667c6f46ce5b357c140b019","abstract_canon_sha256":"534fe76e0dfbf7c293edda4666f9e3c40e246050a72cfa8d9e2bd17bae455c6d"},"schema_version":"1.0"},"canonical_sha256":"85f0c4120391363d461736a064907c2dfd47c8b5b15737f9d1667bc8ee7640bd","receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-05-20T00:02:12.600381Z","signature_b64":"7pnFYRei5uSVpx0CObnEdM7id8615mhLdNpPl2+SEpR06WzFwneT5dyy1Jrp3j+ixhbhPghIuNnlULUOllEJBg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"85f0c4120391363d461736a064907c2dfd47c8b5b15737f9d1667bc8ee7640bd","last_reissued_at":"2026-05-20T00:02:12.599667Z","signature_status":"signed_v1","first_computed_at":"2026-05-20T00:02:12.599667Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"source_kind":"arxiv","source_id":"2605.07022","source_version":2,"attestation_state":"computed"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:02:12Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"4GxczwY0TywFdLS71rZtkoh6eDtC8oYl9P4JprPI//UFRvvtH8ixDxOqGDDFIZoXHS1MSbHTI497aE7usRAAAg==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T21:05:29.384368Z"},"content_sha256":"e362244f0b99cafdb68a5ae8f1a026b9cada2493cb706fc4bb6371a98ab818c3","schema_version":"1.0","event_id":"sha256:e362244f0b99cafdb68a5ae8f1a026b9cada2493cb706fc4bb6371a98ab818c3"},{"event_type":"graph_snapshot","subject_pith_number":"pith:2026:QXYMIEQDSE3D2RQXG2QGJED4FX","target":"graph","payload":{"graph_snapshot":{"paper":{"title":"Self-Driving Datasets: From 20 Million Papers to Nuanced Biomedical Knowledge at Scale","license":"http://creativecommons.org/licenses/by/4.0/","headline":"PubMed can be autonomously turned into structured biomedical datasets larger, more nuanced, and more accurate than the curated databases they replace.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Alden Rose, Cesar de la Fuente-Nunez, Haydn Jones, Jacob R. Gardner, Jiaming Liang, Kaiwen Wu, Li S. Yifei, Maggie Ziyu Huan, Mark Yatskar, Osbert Bastani, Yimeng Zeng, Yining Huang, Yoseph Barash, Zachary Ives","submitted_at":"2026-05-07T23:08:18Z","abstract_excerpt":"Manually curated biomedical repositories -- spanning bioactivity, genomics, and chemistry -- are expensive to maintain, lag behind primary literature, and discard experimental context, obscuring nuances needed to assess data correctness and coverage. We show that PubMed itself can be autonomously and cost-effectively turned into structured datasets that are larger, more nuanced, and more accurate than the curated databases they replace. We present three coupled contributions: (1) an LLM-based entity-tagging pipeline, grounded in nine biomedical ontologies, that tags 4.5B entities across 19 cat"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"PubMed itself can be autonomously and cost-effectively turned into structured datasets that are larger, more nuanced, and more accurate than the curated databases they replace.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"That frontier-model rejection rates on the extracted records provide a reliable proxy for actual correctness and that the multi-agent extraction process does not introduce systematic biases not captured by those checks.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"Starling uses LLMs and agents to turn 22.5M PubMed papers into 6.3M nuanced structured records across six tasks with 0.6-7.7% frontier-model rejection rates, lower than error rates on existing curated databases.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"PubMed can be autonomously turned into structured biomedical datasets larger, more nuanced, and more accurate than the curated databases they replace.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3b2270b612cf989e319686fb08c086106efe7256c4daf28680273d5abb4797ec"},"source":{"id":"2605.07022","kind":"arxiv","version":2},"verdict":{"id":"2a942254-5df8-4af6-8dcc-38ea21c84ab2","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-11T01:23:53.375296Z","strongest_claim":"PubMed itself can be autonomously and cost-effectively turned into structured datasets that are larger, more nuanced, and more accurate than the curated databases they replace.","one_line_summary":"Starling uses LLMs and agents to turn 22.5M PubMed papers into 6.3M nuanced structured records across six tasks with 0.6-7.7% frontier-model rejection rates, lower than error rates on existing curated databases.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"That frontier-model rejection rates on the extracted records provide a reliable proxy for actual correctness and that the multi-agent extraction process does not introduce systematic biases not captured by those checks.","pith_extraction_headline":"PubMed can be autonomously turned into structured biomedical datasets larger, more nuanced, and more accurate than the curated databases they replace."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2605.07022/integrity.json","findings":[],"available":true,"detectors_run":[{"name":"doi_title_agreement","ran_at":"2026-05-19T17:31:18.935057Z","status":"completed","version":"1.0.0","findings_count":0},{"name":"doi_compliance","ran_at":"2026-05-19T12:11:24.882050Z","status":"completed","version":"1.0.0","findings_count":0}],"snapshot_sha256":"44f9bce28235a650fa59fdcc63ec6f8ac1aecf801fe2dfd0e7abff9033670e7d"},"references":{"count":0,"sample":[],"resolved_work":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","internal_anchors":0},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"},"verdict_id":"2a942254-5df8-4af6-8dcc-38ea21c84ab2"},"signer":{"signer_id":"pith.science","signer_type":"pith_registry","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"created_at":"2026-05-20T00:02:12Z","supersedes":[],"prev_event":null,"signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"jTI1TwN/Q6+XZNCRU6D9ro/wCVxf/d3obttKoyT5wVsCvUFyLnih7eFd0Aa2mIWAbOnKZNTW0i6w4vFAHbULAw==","signed_message":"open_graph_event_sha256_bytes","signed_at":"2026-06-02T21:05:29.384894Z"},"content_sha256":"6b29989e66fdfc44ac68f460f311b64118c1e93fafce16e585bf4f1396520a99","schema_version":"1.0","event_id":"sha256:6b29989e66fdfc44ac68f460f311b64118c1e93fafce16e585bf4f1396520a99"}],"timestamp_proofs":[],"mirror_hints":[{"mirror_type":"https","name":"Pith Resolver","base_url":"https://pith.science","bundle_url":"https://pith.science/pith/QXYMIEQDSE3D2RQXG2QGJED4FX/bundle.json","state_url":"https://pith.science/pith/QXYMIEQDSE3D2RQXG2QGJED4FX/state.json","well_known_bundle_url":"https://pith.science/.well-known/pith/QXYMIEQDSE3D2RQXG2QGJED4FX/bundle.json","status":"primary"}],"public_keys":[{"key_id":"pith-v1-2026-05","algorithm":"ed25519","format":"raw","public_key_b64":"stVStoiQhXFxp4s2pdzPNoqVNBMojDU/fJ2db5S3CbM=","public_key_hex":"b2d552b68890857171a78b36a5dccf368a953413288c353f7c9d9d6f94b709b3","fingerprint_sha256_b32_first128bits":"RVFV5Z2OI2J3ZUO7ERDEBCYNKS","fingerprint_sha256_hex":"8d4b5ee74e4693bcd1df2446408b0d54","rotates_at":null,"url":"https://pith.science/pith-signing-key.json","notes":"Pith uses this Ed25519 key to sign canonical record SHA-256 digests. Verify with: ed25519_verify(public_key, message=canonical_sha256_bytes, signature=base64decode(signature_b64))."}],"merge_version":"pith-open-graph-merge-v1","built_at":"2026-06-02T21:05:29Z","links":{"resolver":"https://pith.science/pith/QXYMIEQDSE3D2RQXG2QGJED4FX","bundle":"https://pith.science/pith/QXYMIEQDSE3D2RQXG2QGJED4FX/bundle.json","state":"https://pith.science/pith/QXYMIEQDSE3D2RQXG2QGJED4FX/state.json","well_known_bundle":"https://pith.science/.well-known/pith/QXYMIEQDSE3D2RQXG2QGJED4FX/bundle.json"},"state":{"state_type":"pith_open_graph_state","state_version":"1.0","pith_number":"pith:2026:QXYMIEQDSE3D2RQXG2QGJED4FX","merge_version":"pith-open-graph-merge-v1","event_count":2,"valid_event_count":2,"invalid_event_count":0,"equivocation_count":0,"current":{"canonical_record":{"metadata":{"abstract_canon_sha256":"534fe76e0dfbf7c293edda4666f9e3c40e246050a72cfa8d9e2bd17bae455c6d","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T23:08:18Z","title_canon_sha256":"4b83343cffc37ad48fdb397209199bc51669e3276667c6f46ce5b357c140b019"},"schema_version":"1.0","source":{"id":"2605.07022","kind":"arxiv","version":2}},"source_aliases":[{"alias_kind":"arxiv","alias_value":"2605.07022","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"arxiv_version","alias_value":"2605.07022v2","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2605.07022","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"pith_short_12","alias_value":"QXYMIEQDSE3D","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"pith_short_16","alias_value":"QXYMIEQDSE3D2RQX","created_at":"2026-05-20T00:02:12Z"},{"alias_kind":"pith_short_8","alias_value":"QXYMIEQD","created_at":"2026-05-20T00:02:12Z"}],"graph_snapshots":[{"event_id":"sha256:6b29989e66fdfc44ac68f460f311b64118c1e93fafce16e585bf4f1396520a99","target":"graph","created_at":"2026-05-20T00:02:12Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"graph_snapshot":{"author_claims":{"count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57","strong_count":0},"builder_version":"pith-number-builder-2026-05-17-v1","claims":{"count":4,"items":[{"attestation":"unclaimed","claim_id":"C1","kind":"strongest_claim","source":"verdict.strongest_claim","status":"machine_extracted","text":"PubMed itself can be autonomously and cost-effectively turned into structured datasets that are larger, more nuanced, and more accurate than the curated databases they replace."},{"attestation":"unclaimed","claim_id":"C2","kind":"weakest_assumption","source":"verdict.weakest_assumption","status":"machine_extracted","text":"That frontier-model rejection rates on the extracted records provide a reliable proxy for actual correctness and that the multi-agent extraction process does not introduce systematic biases not captured by those checks."},{"attestation":"unclaimed","claim_id":"C3","kind":"one_line_summary","source":"verdict.one_line_summary","status":"machine_extracted","text":"Starling uses LLMs and agents to turn 22.5M PubMed papers into 6.3M nuanced structured records across six tasks with 0.6-7.7% frontier-model rejection rates, lower than error rates on existing curated databases."},{"attestation":"unclaimed","claim_id":"C4","kind":"headline","source":"verdict.pith_extraction.headline","status":"machine_extracted","text":"PubMed can be autonomously turned into structured biomedical datasets larger, more nuanced, and more accurate than the curated databases they replace."}],"snapshot_sha256":"3b2270b612cf989e319686fb08c086106efe7256c4daf28680273d5abb4797ec"},"formal_canon":{"evidence_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"integrity":{"available":true,"clean":true,"detectors_run":[{"findings_count":0,"name":"doi_title_agreement","ran_at":"2026-05-19T17:31:18.935057Z","status":"completed","version":"1.0.0"},{"findings_count":0,"name":"doi_compliance","ran_at":"2026-05-19T12:11:24.882050Z","status":"completed","version":"1.0.0"}],"endpoint":"/pith/2605.07022/integrity.json","findings":[],"snapshot_sha256":"44f9bce28235a650fa59fdcc63ec6f8ac1aecf801fe2dfd0e7abff9033670e7d","summary":{"advisory":0,"by_detector":{},"critical":0,"informational":0}},"paper":{"abstract_excerpt":"Manually curated biomedical repositories -- spanning bioactivity, genomics, and chemistry -- are expensive to maintain, lag behind primary literature, and discard experimental context, obscuring nuances needed to assess data correctness and coverage. We show that PubMed itself can be autonomously and cost-effectively turned into structured datasets that are larger, more nuanced, and more accurate than the curated databases they replace. We present three coupled contributions: (1) an LLM-based entity-tagging pipeline, grounded in nine biomedical ontologies, that tags 4.5B entities across 19 cat","authors_text":"Alden Rose, Cesar de la Fuente-Nunez, Haydn Jones, Jacob R. Gardner, Jiaming Liang, Kaiwen Wu, Li S. Yifei, Maggie Ziyu Huan, Mark Yatskar, Osbert Bastani, Yimeng Zeng, Yining Huang, Yoseph Barash, Zachary Ives","cross_cats":[],"headline":"PubMed can be autonomously turned into structured biomedical datasets larger, more nuanced, and more accurate than the curated databases they replace.","license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T23:08:18Z","title":"Self-Driving Datasets: From 20 Million Papers to Nuanced Biomedical Knowledge at Scale"},"references":{"count":0,"internal_anchors":0,"resolved_work":0,"sample":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2605.07022","kind":"arxiv","version":2},"verdict":{"created_at":"2026-05-11T01:23:53.375296Z","id":"2a942254-5df8-4af6-8dcc-38ea21c84ab2","model_set":{"reader":"grok-4.3"},"one_line_summary":"Starling uses LLMs and agents to turn 22.5M PubMed papers into 6.3M nuanced structured records across six tasks with 0.6-7.7% frontier-model rejection rates, lower than error rates on existing curated databases.","pipeline_version":"pith-pipeline@v0.9.0","pith_extraction_headline":"PubMed can be autonomously turned into structured biomedical datasets larger, more nuanced, and more accurate than the curated databases they replace.","strongest_claim":"PubMed itself can be autonomously and cost-effectively turned into structured datasets that are larger, more nuanced, and more accurate than the curated databases they replace.","weakest_assumption":"That frontier-model rejection rates on the extracted records provide a reliable proxy for actual correctness and that the multi-agent extraction process does not introduce systematic biases not captured by those checks."}},"verdict_id":"2a942254-5df8-4af6-8dcc-38ea21c84ab2"}}],"author_attestations":[],"timestamp_anchors":[],"storage_attestations":[],"citation_signatures":[],"replication_records":[],"corrections":[],"mirror_hints":[],"record_created":{"event_id":"sha256:e362244f0b99cafdb68a5ae8f1a026b9cada2493cb706fc4bb6371a98ab818c3","target":"record","created_at":"2026-05-20T00:02:12Z","signer":{"key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signer_id":"pith.science","signer_type":"pith_registry"},"payload":{"attestation_state":"computed","canonical_record":{"metadata":{"abstract_canon_sha256":"534fe76e0dfbf7c293edda4666f9e3c40e246050a72cfa8d9e2bd17bae455c6d","cross_cats_sorted":[],"license":"http://creativecommons.org/licenses/by/4.0/","primary_cat":"cs.LG","submitted_at":"2026-05-07T23:08:18Z","title_canon_sha256":"4b83343cffc37ad48fdb397209199bc51669e3276667c6f46ce5b357c140b019"},"schema_version":"1.0","source":{"id":"2605.07022","kind":"arxiv","version":2}},"canonical_sha256":"85f0c4120391363d461736a064907c2dfd47c8b5b15737f9d1667bc8ee7640bd","receipt":{"algorithm":"ed25519","builder_version":"pith-number-builder-2026-05-17-v1","canonical_sha256":"85f0c4120391363d461736a064907c2dfd47c8b5b15737f9d1667bc8ee7640bd","first_computed_at":"2026-05-20T00:02:12.599667Z","key_id":"pith-v1-2026-05","kind":"pith_receipt","last_reissued_at":"2026-05-20T00:02:12.599667Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","receipt_version":"0.3","signature_b64":"7pnFYRei5uSVpx0CObnEdM7id8615mhLdNpPl2+SEpR06WzFwneT5dyy1Jrp3j+ixhbhPghIuNnlULUOllEJBg==","signature_status":"signed_v1","signed_at":"2026-05-20T00:02:12.600381Z","signed_message":"canonical_sha256_bytes"},"source_id":"2605.07022","source_kind":"arxiv","source_version":2}}},"equivocations":[],"invalid_events":[],"applied_event_ids":["sha256:e362244f0b99cafdb68a5ae8f1a026b9cada2493cb706fc4bb6371a98ab818c3","sha256:6b29989e66fdfc44ac68f460f311b64118c1e93fafce16e585bf4f1396520a99"],"state_sha256":"d8828c1319318c421f5db8d6395ef25e93e290f48420e00cabde2e72b3c598f4"},"bundle_signature":{"signature_status":"signed_v1","algorithm":"ed25519","key_id":"pith-v1-2026-05","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54","signature_b64":"Gl1FuHWW2UoXrzWUeRLUQsBkkWevhd/oviB8J96Xmub35N/JBggd2+vFObKv8/xsviiA8Qsf9HZmYO90ZLTfAA==","signed_message":"bundle_sha256_bytes","signed_at":"2026-06-02T21:05:29.387410Z","bundle_sha256":"f9f0711f706f49d15da226419dd8e9a2b697cbacb8c3e0c8dfcb261dc7379681"}}