{"record_type":"pith_number_record","schema_url":"https://pith.science/schemas/pith-number/v1.json","pith_number":"pith:2025:KYX3QDW73NG3UGGYEXPBXUZT4E","short_pith_number":"pith:KYX3QDW7","schema_version":"1.0","canonical_sha256":"562fb80edfdb4dba18d825de1bd333e1128182fe2fc9d8983a25f35475dc9a4b","source":{"kind":"arxiv","id":"2507.03971","version":1},"attestation_state":"computed","paper":{"title":"Real-TabPFN: Improving Tabular Foundation Models via Continued Pre-training With Real-World Data","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","stat.ME","stat.ML"],"primary_cat":"cs.LG","authors_text":"Anurag Garg, Frank Hutter, Lennart Purucker, Muhammad Ali, Noah Hollmann, Samuel M\\\"uller","submitted_at":"2025-07-05T09:39:07Z","abstract_excerpt":"Foundation models for tabular data, like TabPFN, achieve strong performance on small datasets when pre-trained solely on synthetic data. We show that this performance can be significantly boosted by a targeted continued pre-training phase. Specifically, we demonstrate that leveraging a small, curated collection of large, real-world datasets for continued pre-training yields superior downstream predictive accuracy compared to using broader, potentially noisier corpora like CommonCrawl or GitTables. Our resulting model, Real-TabPFN, achieves substantial performance gains on 29 datasets from the "},"verification_status":{"content_addressed":true,"pith_receipt":true,"author_attested":false,"weak_author_claims":0,"strong_author_claims":0,"externally_anchored":false,"storage_verified":false,"citation_signatures":0,"replication_records":0,"graph_snapshot":true,"references_resolved":false,"formal_links_present":false},"canonical_record":{"source":{"id":"2507.03971","kind":"arxiv","version":1},"metadata":{"license":"http://creativecommons.org/licenses/by-sa/4.0/","primary_cat":"cs.LG","submitted_at":"2025-07-05T09:39:07Z","cross_cats_sorted":["cs.AI","stat.ME","stat.ML"],"title_canon_sha256":"9a7b89cc342eb194a652beae2352820f310ef9b51820c0482c14cf9bc62a9444","abstract_canon_sha256":"ee2f254f0b1397ae21fca3b81bbfaed8f9db8f7d6d22befd8ac7a3aa8b47b792"},"schema_version":"1.0"},"receipt":{"kind":"pith_receipt","key_id":"pith-v1-2026-05","algorithm":"ed25519","signed_at":"2026-07-05T11:32:38.218018Z","signature_b64":"amc/Rv4lM7Mf6fJUl1xfu7zRxV5bk+Nn2nt8jCc7kLTkp/YxhVmg1ZJkfCYwU8NDSbVB8TkoRC204ZSD++y1Cg==","signed_message":"canonical_sha256_bytes","builder_version":"pith-number-builder-2026-05-17-v1","receipt_version":"0.3","canonical_sha256":"562fb80edfdb4dba18d825de1bd333e1128182fe2fc9d8983a25f35475dc9a4b","last_reissued_at":"2026-07-05T11:32:38.217293Z","signature_status":"signed_v1","first_computed_at":"2026-07-05T11:32:38.217293Z","public_key_fingerprint":"8d4b5ee74e4693bcd1df2446408b0d54"},"graph_snapshot":{"paper":{"title":"Real-TabPFN: Improving Tabular Foundation Models via Continued Pre-training With Real-World Data","license":"http://creativecommons.org/licenses/by-sa/4.0/","headline":"","cross_cats":["cs.AI","stat.ME","stat.ML"],"primary_cat":"cs.LG","authors_text":"Anurag Garg, Frank Hutter, Lennart Purucker, Muhammad Ali, Noah Hollmann, Samuel M\\\"uller","submitted_at":"2025-07-05T09:39:07Z","abstract_excerpt":"Foundation models for tabular data, like TabPFN, achieve strong performance on small datasets when pre-trained solely on synthetic data. We show that this performance can be significantly boosted by a targeted continued pre-training phase. Specifically, we demonstrate that leveraging a small, curated collection of large, real-world datasets for continued pre-training yields superior downstream predictive accuracy compared to using broader, potentially noisier corpora like CommonCrawl or GitTables. Our resulting model, Real-TabPFN, achieves substantial performance gains on 29 datasets from the "},"claims":{"count":0,"items":[],"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"source":{"id":"2507.03971","kind":"arxiv","version":1},"verdict":{"id":null,"model_set":{},"created_at":null,"strongest_claim":"","one_line_summary":"","pipeline_version":null,"weakest_assumption":"","pith_extraction_headline":""},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2507.03971/integrity.json","findings":[],"available":true,"detectors_run":[],"snapshot_sha256":"c28c3603d3b5d939e8dc4c7e95fa8dfce3d595e45f758748cecf8e644a296938"},"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"},"aliases":[{"alias_kind":"arxiv","alias_value":"2507.03971","created_at":"2026-07-05T11:32:38.217406+00:00"},{"alias_kind":"arxiv_version","alias_value":"2507.03971v1","created_at":"2026-07-05T11:32:38.217406+00:00"},{"alias_kind":"doi","alias_value":"10.48550/arxiv.2507.03971","created_at":"2026-07-05T11:32:38.217406+00:00"},{"alias_kind":"pith_short_12","alias_value":"KYX3QDW73NG3","created_at":"2026-07-05T11:32:38.217406+00:00"},{"alias_kind":"pith_short_16","alias_value":"KYX3QDW73NG3UGGY","created_at":"2026-07-05T11:32:38.217406+00:00"},{"alias_kind":"pith_short_8","alias_value":"KYX3QDW7","created_at":"2026-07-05T11:32:38.217406+00:00"}],"events":[],"event_summary":{},"paper_claims":[],"inbound_citations":{"count":12,"internal_anchor_count":0,"sample":[{"citing_arxiv_id":"2606.25197","citing_title":"Efficient Adaptive Data Acquisition via Pretrained Belief Representations","ref_index":18,"is_internal_anchor":false},{"citing_arxiv_id":"2606.31208","citing_title":"Probing Memorization of Tabular In-Context Learning","ref_index":169,"is_internal_anchor":false},{"citing_arxiv_id":"2606.07674","citing_title":"Simultaneous hyperkinetic movement disorders phenotyping: a cross-cohort pediatric transfer study using routine videos, markerless pose estimation and a tabular foundation model","ref_index":17,"is_internal_anchor":false},{"citing_arxiv_id":"2603.16513","citing_title":"FEAT: A Linear-Complexity Foundation Model for Extremely Large Structured Data","ref_index":12,"is_internal_anchor":false},{"citing_arxiv_id":"2605.18696","citing_title":"Ensembling Tabular Foundation Models - A Diversity Ceiling And A Calibration Trap","ref_index":9,"is_internal_anchor":false},{"citing_arxiv_id":"2605.19662","citing_title":"When Tabular Foundation Models Meet Strategic Tabular Data: A Prior Alignment Approach","ref_index":88,"is_internal_anchor":false},{"citing_arxiv_id":"2506.16791","citing_title":"TabArena: A Living Benchmark for Machine Learning on Tabular Data","ref_index":127,"is_internal_anchor":false},{"citing_arxiv_id":"2602.09329","citing_title":"MacrOData: New Benchmarks of Thousands of Datasets for Tabular Outlier Detection","ref_index":23,"is_internal_anchor":false},{"citing_arxiv_id":"2511.08667","citing_title":"TabPFN-2.5: Advancing the State of the Art in Tabular Foundation Models","ref_index":19,"is_internal_anchor":false},{"citing_arxiv_id":"2604.12596","citing_title":"KumoRFM-2: Scaling Foundation Models for Relational Learning","ref_index":5,"is_internal_anchor":false},{"citing_arxiv_id":"2604.16123","citing_title":"Tabular foundation models for in-context prediction of molecular properties","ref_index":43,"is_internal_anchor":false},{"citing_arxiv_id":"2605.04911","citing_title":"Breaking the Quality-Privacy Tradeoff in Tabular Data Generation via In-Context Learning","ref_index":9,"is_internal_anchor":false}]},"formal_canon":{"evidence_count":0,"sample":[],"anchors":[]},"links":{"html":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E","json":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E.json","graph_json":"https://pith.science/api/pith-number/KYX3QDW73NG3UGGYEXPBXUZT4E/graph.json","events_json":"https://pith.science/api/pith-number/KYX3QDW73NG3UGGYEXPBXUZT4E/events.json","paper":"https://pith.science/paper/KYX3QDW7"},"agent_actions":{"view_html":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E","download_json":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E.json","view_paper":"https://pith.science/paper/KYX3QDW7","resolve_alias":"https://pith.science/api/pith-number/resolve?arxiv=2507.03971&json=true","fetch_graph":"https://pith.science/api/pith-number/KYX3QDW73NG3UGGYEXPBXUZT4E/graph.json","fetch_events":"https://pith.science/api/pith-number/KYX3QDW73NG3UGGYEXPBXUZT4E/events.json","actions":{"anchor_timestamp":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E/action/timestamp_anchor","attest_storage":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E/action/storage_attestation","attest_author":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E/action/author_attestation","sign_citation":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E/action/citation_signature","submit_replication":"https://pith.science/pith/KYX3QDW73NG3UGGYEXPBXUZT4E/action/replication_record"}},"created_at":"2026-07-05T11:32:38.217406+00:00","updated_at":"2026-07-05T11:32:38.217406+00:00"}