Attention in minimal transformers under corruption performs in-context empirical Bayes via a single kernel-weighted posterior mean step followed by depth-driven particle dynamics refinement.
Solving empirical Bayes via transform- ers
4 Pith papers cite this work. Polarity classification is still indexing.
4
Pith papers citing it
years
2026 4representative citing papers
A Gamma-smoothed NPMLE for Poisson empirical Bayes achieves optimal nearly parametric rates for posterior means and enables asymptotically exact, shorter marginal coverage confidence sets under compact support.
Dirichlet-process Bayes and Newton quasi-Bayes g-modeling procedures for the Poisson compound decision problem merge in Hellinger distance and regret at rates controlled by the learning-rate exponent, with a multidimensional extension.
InfoAtlas is a pretrained neural model for zero-shot mutual information estimation that matches state-of-the-art accuracy with 100x speedup and handles varying dimensions via a single model.
citing papers explorer
-
Attention as In-Context Empirical Bayes: A Two-Stage View via Particle Dynamics
Attention in minimal transformers under corruption performs in-context empirical Bayes via a single kernel-weighted posterior mean step followed by depth-driven particle dynamics refinement.
-
Poisson Empirical Bayes via Gamma-Smoothed Nonparametric Maximum Likelihood
A Gamma-smoothed NPMLE for Poisson empirical Bayes achieves optimal nearly parametric rates for posterior means and enables asymptotically exact, shorter marginal coverage confidence sets under compact support.
-
Merging of Bayes and quasi-Bayes empirical Bayes procedures for Poisson compound decisions
Dirichlet-process Bayes and Newton quasi-Bayes g-modeling procedures for the Poisson compound decision problem merge in Hellinger distance and regret at rates controlled by the learning-rate exponent, with a multidimensional extension.
-
InfoAtlas: A Foundation Model for Zero-Shot Statistical Dependence Estimate
InfoAtlas is a pretrained neural model for zero-shot mutual information estimation that matches state-of-the-art accuracy with 100x speedup and handles varying dimensions via a single model.