{"paper":{"title":"Relational In-Context Learning via Synthetic Pre-training with Structural Prior","license":"http://arxiv.org/licenses/nonexclusive-distrib/1.0/","headline":"RDB-PFN learns relational in-context adaptation by pre-training a transformer solely on millions of synthetic databases generated from structural causal models.","cross_cats":["cs.AI","cs.DB"],"primary_cat":"cs.LG","authors_text":"Chuan Shi, Jiaxuan You, Muhan Zhang, Yanbo Wang","submitted_at":"2026-03-04T07:30:54Z","abstract_excerpt":"Relational Databases (RDBs) are the backbone of modern business, yet they lack foundation models comparable to those in text or vision. A key obstacle is that high-quality RDBs are private, scarce, and structurally heterogeneous, making internet-scale pre-training infeasible. To overcome this data scarcity, we introduce RDB-PFN, the first relational foundation model trained purely via synthetic data. Inspired by Prior-Data Fitted Networks (PFNs), where synthetic data generated from Structural Causal Models (SCMs) enables reasoning on single tables, we design a Relational Prior Generator to cre"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"RDB-PFN achieves strong few-shot performance on 19 real-world relational prediction tasks, outperforming graph-based and single-table foundation-model baselines (given the same DFS-linearized inputs), while using a lightweight architecture and fast inference.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"Synthetic relational databases generated by the Relational Prior Generator from Structural Causal Models sufficiently capture the structural heterogeneity, join patterns, and statistical properties of real-world RDBs to support generalization to actual tasks.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"RDB-PFN is a relational foundation model pre-trained on over 2 million synthetic RDB tasks that achieves strong few-shot performance on 19 real-world relational prediction tasks via in-context learning.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"RDB-PFN learns relational in-context adaptation by pre-training a transformer solely on millions of synthetic databases generated from structural causal models.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"731a49ddb63e75d60ae7f450fa896cbc87a63ffad9506e086d806e561857c4c0"},"source":{"id":"2603.03805","kind":"arxiv","version":5},"verdict":{"id":"15c12862-40cd-43e1-ba34-e4111618b7d7","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-15T16:54:03.121535Z","strongest_claim":"RDB-PFN achieves strong few-shot performance on 19 real-world relational prediction tasks, outperforming graph-based and single-table foundation-model baselines (given the same DFS-linearized inputs), while using a lightweight architecture and fast inference.","one_line_summary":"RDB-PFN is a relational foundation model pre-trained on over 2 million synthetic RDB tasks that achieves strong few-shot performance on 19 real-world relational prediction tasks via in-context learning.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"Synthetic relational databases generated by the Relational Prior Generator from Structural Causal Models sufficiently capture the structural heterogeneity, join patterns, and statistical properties of real-world RDBs to support generalization to actual tasks.","pith_extraction_headline":"RDB-PFN learns relational in-context adaptation by pre-training a transformer solely on millions of synthetic databases generated from structural causal models."},"integrity":{"clean":true,"summary":{"advisory":0,"critical":0,"by_detector":{},"informational":0},"endpoint":"/pith/2603.03805/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":2,"snapshot_sha256":"0061bd0a8b837f3790f2ef59463bafa6c7978bf7d4531bada606f0242f22c935"},"author_claims":{"count":0,"strong_count":0,"snapshot_sha256":"258153158e38e3291e3d48162225fcdb2d5a3ed65a07baac614ab91432fd4f57"},"builder_version":"pith-number-builder-2026-05-17-v1"}