TabMGP: Martingale Posterior with TabPFN
Fuente:
arXiv
Saved in:
| Main Authors: | , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866913170262589440 |
|---|---|
| author | Ng, Kenyon Fong, Edwin Frazier, David T. Knoblauch, Jeremias Wei, Susan |
| author_facet | Ng, Kenyon Fong, Edwin Frazier, David T. Knoblauch, Jeremias Wei, Susan |
| contents | Bayesian inference provides principled uncertainty quantification but is often limited by the challenges of prior and likelihood elicitation. The martingale posterior (MGP) (Fong et al., 2023) offers an alternative by replacing these requirements with a predictive rule. In addition, the MGP focuses inference on parameters defined through a loss function. This framework is especially resonant in the era of foundation transformers; practitioners increasingly leverage models like TabPFN for their state-of-the-art capabilities, yet often require epistemic uncertainty for a scientific estimand $θ$ that need not parameterise the implicit latent model. The MGP provides a mechanism to recover these posterior distributions. We introduce TabMGP, an MGP built on TabPFN for tabular data. TabMGP produces credible sets with near-nominal coverage and often outperforms both handcrafted MGP constructions and standard Bayesian baselines. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2510_25154 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | TabMGP: Martingale Posterior with TabPFN Ng, Kenyon Fong, Edwin Frazier, David T. Knoblauch, Jeremias Wei, Susan Methodology Computation Machine Learning 62F15 Bayesian inference provides principled uncertainty quantification but is often limited by the challenges of prior and likelihood elicitation. The martingale posterior (MGP) (Fong et al., 2023) offers an alternative by replacing these requirements with a predictive rule. In addition, the MGP focuses inference on parameters defined through a loss function. This framework is especially resonant in the era of foundation transformers; practitioners increasingly leverage models like TabPFN for their state-of-the-art capabilities, yet often require epistemic uncertainty for a scientific estimand $θ$ that need not parameterise the implicit latent model. The MGP provides a mechanism to recover these posterior distributions. We introduce TabMGP, an MGP built on TabPFN for tabular data. TabMGP produces credible sets with near-nominal coverage and often outperforms both handcrafted MGP constructions and standard Bayesian baselines. |
| title | TabMGP: Martingale Posterior with TabPFN |
| topic | Methodology Computation Machine Learning 62F15 |
| url | https://arxiv.org/abs/2510.25154 |