{"paper":{"title":"TabClustPFN: A Prior-Fitted Network for Tabular Data Clustering","license":"http://creativecommons.org/licenses/by/4.0/","headline":"TabClustPFN clusters any new tabular dataset in one forward pass by amortizing Bayesian inference over assignments and cluster count.","cross_cats":[],"primary_cat":"cs.LG","authors_text":"Guanyang Wang, Qiong Zhang, Tianqi Zhao, Yan Shuo Tan","submitted_at":"2026-01-29T12:56:41Z","abstract_excerpt":"Clustering tabular data is a fundamental yet challenging problem due to heterogeneous feature types, diverse data-generating mechanisms, and the absence of transferable inductive biases across datasets. Prior-fitted networks (PFNs) have recently demonstrated strong generalization in supervised tabular learning by amortizing Bayesian inference under a broad synthetic prior. Extending this paradigm to clustering is nontrivial: clustering is unsupervised, admits a combinatorial and permutation-invariant output space, and requires inferring the number of clusters. We introduce TabClustPFN, a prior"},"claims":{"count":4,"items":[{"kind":"strongest_claim","text":"TabClustPFN clusters unseen datasets in a single forward pass, without dataset-specific retraining or hyperparameter tuning, and outperforms classical, deep, and amortized clustering baselines on synthetic and real-world tabular benchmarks.","source":"verdict.strongest_claim","status":"machine_extracted","claim_id":"C1","attestation":"unclaimed"},{"kind":"weakest_assumption","text":"The synthetic datasets drawn from the flexible clustering prior sufficiently resemble the structure and heterogeneity of real-world tabular data so that the pretrained network generalizes without retraining.","source":"verdict.weakest_assumption","status":"machine_extracted","claim_id":"C2","attestation":"unclaimed"},{"kind":"one_line_summary","text":"TabClustPFN performs amortized Bayesian inference for cluster assignments and cardinality on unseen tabular data after pretraining on synthetic data from a flexible prior.","source":"verdict.one_line_summary","status":"machine_extracted","claim_id":"C3","attestation":"unclaimed"},{"kind":"headline","text":"TabClustPFN clusters any new tabular dataset in one forward pass by amortizing Bayesian inference over assignments and cluster count.","source":"verdict.pith_extraction.headline","status":"machine_extracted","claim_id":"C4","attestation":"unclaimed"}],"snapshot_sha256":"3610ef08aff7512d9ad37ada2a4ca7d20ce55658998676b9a0ca8e3a29fe09c2"},"source":{"id":"2601.21656","kind":"arxiv","version":3},"verdict":{"id":"60e87dcc-2129-489e-9d70-363a3a0daeab","model_set":{"reader":"grok-4.3"},"created_at":"2026-05-16T10:28:09.541684Z","strongest_claim":"TabClustPFN clusters unseen datasets in a single forward pass, without dataset-specific retraining or hyperparameter tuning, and outperforms classical, deep, and amortized clustering baselines on synthetic and real-world tabular benchmarks.","one_line_summary":"TabClustPFN performs amortized Bayesian inference for cluster assignments and cardinality on unseen tabular data after pretraining on synthetic data from a flexible prior.","pipeline_version":"pith-pipeline@v0.9.0","weakest_assumption":"The synthetic datasets drawn from the flexible clustering prior sufficiently resemble the structure and heterogeneity of real-world tabular data so that the pretrained network generalizes without retraining.","pith_extraction_headline":"TabClustPFN clusters any new tabular dataset in one forward pass by amortizing Bayesian inference over assignments and cluster count."},"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"}