Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
TabMGP: Martingale Posterior with TabPFN
Kenyon Ng, Edwin Fong, David Tyler Frazier, Jeremias Knoblauch, Susan Wei
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use TabPFN as predictive rule instead of Bayesian bootstrap | Martingale posterior · linear regression, synthetic setups | coverage: 0.55 -> 1.00 (N(0,1)) | Improved |
| use TabPFN predictive rule instead of Gaussian copula update | Martingale posterior · linear regression, kin8nm | coverage: 0.63 -> 0.95 | Improved |
| use TabMGP instead of asymptotic Gaussian | linear regression credible sets · synthetic linear regression, N(0,1) | credible set size: 1.31 -> 0.45 | Improved |
| use TabMGP instead of Asymptotic | linear regression credible sets · abalone | credible set size: 1.64 -> 0.95 | Improved |
| use TabMGP instead of Asymptotic in synthetic logistic regression | logistic regression credible sets · synthetic logistic link | coverage: 0.96 -> 0.84 | Hurt |
| use TabMGP instead of Bayes | multinomial logistic regression · wine | coverage: 0.09 -> 0.93 | Improved |
| use TabMGP instead of BB | multinomial logistic regression · yeast | coverage: 0.68 -> 0.97 | Improved |
| use TabMGP instead of Copula | binary logistic regression · telescope | credible set size: 50.69 -> 5.96 | Improved |
| use TabMGP in small n/p classification | TabMGP · synthetic logistic regression | coverage: 0.78 (GMM(-2)) vs 0.95 target | Hurt |
| increase rollout length N to 500+ forward steps | TabMGP · 30 setups | convergence of theta(F_N) | Improved |
Closest papers by meaning
- Effortless, Simulation-Efficient Bayesian Inference using Tabular Foundation Models (NeurIPS 2025 · similarity 0.86)
- Asymptotically Log-Optimal Bayes-Assisted Confidence Sequences for Bounded Means (NeurIPS 2026 · similarity 0.84)
- Drift-Resilient TabPFN: In-Context Learning Temporal Distribution Shifts on Tabular Data (NeurIPS 2024 · similarity 0.84)
- Martingale Posterior Neural Networks for Fast Sequential Decision Making (NeurIPS 2025 · similarity 0.83)
- Learning Likelihood-Free Reference Priors (ICML 2025 · similarity 0.83)
- Model-Informed Flows for Bayesian Inference (NeurIPS 2025 · similarity 0.82)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
