TabDPT in-context tabular-foundation discriminator (conditional single-sample)

Conditional-single-sample discriminator (family O) built on TabDPT, the pretrained tabular foundation model of Ma et al. (2024). Like TabPFN and TabICL, TabDPT classifies a query row by in-context learning: a forward pass of a frozen pretrained transformer conditions on a provided labelled TRAINING context (optionally retrieval-augmented) and emits the query's class posterior. There is no per-dataset gradient training – "fitting" only loads the training context into the model. This scorer is the direct sibling of fit_tabpfn and fit_tabicl, swapping the in-context forward calls for TabDPT's.

Why this is single-sample-deployable (conditional-in-N). A specimen is single-sample-classifiable iff the TRAINING context that conditions the transformer is FROZEN into the model at fit time, so that at inference the query's posterior depends only on that query row and the frozen context – never on which other specimens are scored alongside it. This implementation freezes the per-sample-rCLR-transformed training context (rows + labels) and the feature universe at fit, and at score conditions every query against that one frozen context. The frozen context is the TRAINING rows only, so it is leakage-safe (no scored / test data ever enters the context). The scorer pins the context-reduction mode to "subsample" (a single context shared across queries, not per-query retrieval) and passes the full frozen context (no per-query subsampling), so the conditioning set is identical for every query and never depends on the scored batch.

Row-independent scoring modes. With the frozen shared (subsample) context and the full context passed to every query, TabDPT's predict_proba is invariant to the row ORDER of the scored batch and to the scored subset – a query's posterior does not depend on the other queries (measured companion maxdiff \(0\)). The default deployment path still evaluates predict_proba ONE QUERY ROW AT A TIME, so the \(n=1\) forward path is used uniformly. The optional benchmark-only score_batch path evaluates balanced chunks of query rows. That batched path is leakage-free and AUC-faithful (\(|dAUC| = 0\); specimen ranks/verdicts unchanged), but it is not bit-identical to the \(n=1\) row-by-row scorer: the installed build has a deterministic batch-size numerical effect (observed max|b-r| = 1.97e-3), not a cross-row coupling.

Single-sample transform. Each specimen is mapped to the self-contained per-sample robust CLR (rCLR) over its OWN strictly-positive support (geometric-mean centring on v > 0) in the FROZEN feature universe; this is exactly invariant to per-specimen scaling and uses no cross-row statistic. The SAME rCLR transform is applied to the frozen training context at fit and to each query at score. Universe features absent from a specimen carry the neutral rCLR value 0 (centred log-ratio of an absent coordinate), matching the missing-feature rule of the other roster scorers.

Token-free pinned checkpoint. The default TabDPT checkpoint (tabdpt1_2.safetensors, the TabDPT v1.2 model of Ma et al. 2024) is hosted in the public Hugging Face repository Layer6/TabDPT and is fetched non-interactively by hf_hub_download – there is no license acceptance or interactive token step (unlike TabPFN's gated V2.6). This method pins that checkpoint version explicitly (via the installed tabdpt package's pinned _VERSION = "1_2") so the model is reproducible and the download path is token-free.

Orientation and output. The score is the frozen-context posterior \(P(\mathrm{case}) = P(y = 1)\) for the query, in \([0, 1]\); larger = more case-like. Degenerate queries (fewer than hp$min_features universe features present, or an all-zero / empty specimen) return the neutral probability 0.5.

References

Ma J, Thomas V, Hosseinzadeh R, Kamkari H, Labach A, Cresswell JC, Golestan S, Yu G, Volkovs M, Caterini AL. (2024) TabDPT: Scaling Tabular Foundation Models. arXiv:2410.18164.

Qu J, Holzmuller D, Varoquaux G, Le Morvan M. (2025) TabICL: A Tabular Foundation Model for In-Context Learning on Large Data. International Conference on Machine Learning (ICML); arXiv:2502.05564.

Hollmann N, Muller S, Purucker L, et al. (2025) Accurate predictions on small data with a tabular foundation model. Nature 637:319-326.