Federated martingale posterior sampling lets clients share data embeddings for central predictive Bayesian sampling, matching centralized performance and improving calibration on MNIST, CIFAR-10, and CIFAR-100.
Title resolution pending
8 Pith papers cite this work. Polarity classification is still indexing.
8
Pith papers citing it
representative citing papers
Kernel density estimator and recursive kernel predictive processes converge weakly almost surely, with the classic version limiting to compact support and the recursive version to non-compact support.
Variational predictive resampling iteratively imputes data from a variational predictive to produce posterior samples that converge to the exact Bayesian posterior in Gaussian models where mean-field VI retains a gap.
A nonparametric quasi-Bayes empirical Bayes procedure is proposed for estimating sums of random variables, with recursive mixing distribution estimation, asymptotic guarantees, and uncertainty quantification.
Temporal diversity in task distribution during training increases generalization bias over memorization in transformers for in-context linear regression.
Predictive Bayesian inference posteriors concentrate onto a forward-model-dependent quantity and produce miscalibrated credible sets unless the predictive model contains the true data-generating process.
Exchangeable MVPS have Dirichlet process mixture priors tied to emergent conditioning sigma-algebras, with null-component extensions and c.i.d. equivalence for balanced cases.
In multiple-choice QA, LLM beliefs drift early under repeated sampling but self-stabilize; seed-answer prompting and a self-consistency loss reduce drift while preserving accuracy.
citing papers explorer
-
Federated Martingale Posterior Samping
Federated martingale posterior sampling lets clients share data embeddings for central predictive Bayesian sampling, matching centralized performance and improving calibration on MNIST, CIFAR-10, and CIFAR-100.
-
Predictive Inference via Kernel Density Estimates
Kernel density estimator and recursive kernel predictive processes converge weakly almost surely, with the classic version limiting to compact support and the recursive version to non-compact support.
-
Variational predictive resampling
Variational predictive resampling iteratively imputes data from a variational predictive to produce posterior samples that converge to the exact Bayesian posterior in Gaussian models where mean-field VI retains a gap.
-
Quasi-Bayes empirical Bayes estimation of sums of random variables
A nonparametric quasi-Bayes empirical Bayes procedure is proposed for estimating sums of random variables, with recursive mixing distribution estimation, asymptotic guarantees, and uncertainty quantification.
-
Temporal Task Diversity: Inductive Biases Under Non-Stationarity in Synthetic Sequence Modelling
Temporal diversity in task distribution during training increases generalization bias over memorization in transformers for in-context linear regression.
-
Concentration and Calibration in Predictive Bayesian Inference
Predictive Bayesian inference posteriors concentrate onto a forward-model-dependent quantity and produce miscalibrated credible sets unless the predictive model contains the true data-generating process.
-
Some developments of exchangeable measure-valued P\'{o}lya sequences
Exchangeable MVPS have Dirichlet process mixture priors tied to emergent conditioning sigma-algebras, with null-component extensions and c.i.d. equivalence for balanced cases.
-
From Drift to Coherence: Stabilizing Beliefs in LLMs
In multiple-choice QA, LLM beliefs drift early under repeated sampling but self-stabilize; seed-answer prompting and a self-consistency loss reduce drift while preserving accuracy.