PuSH - Publication Server of Helmholtz Zentrum München

Reuter, A.* ; Rudner, T.G.J.* ; Fortuin, V. ; Rügamer, D.*

Can Transformers Learn Full Bayesian Inference In Context?

In: (42nd International Conference on Machine Learning, ICML 2025, 13-19 July 2025, Vancouver). 2025. 51531-51582 (Proceedings of Machine Learning Research ; 267)
Publ. Version/Full Text
Open Access Hybrid
Transformers have emerged as the dominant architecture in the field of deep learning, with a broad range of applications and remarkable in-context learning (ICL) capabilities. While not yet fully understood, ICL has already proved to be an intriguing phenomenon, allowing transformers to learn in context—without requiring further training. In this paper, we further advance the understanding of ICL by demonstrating that transformers can perform full Bayesian inference for commonly used statistical models in context. More specifically, we introduce a general framework that builds on ideas from prior fitted networks and continuous normalizing flows and enables us to infer complex posterior distributions for models such as generalized linear models and latent factor models. Extensive experiments on real-world datasets demonstrate that our ICL approach yields posterior samples that are similar in quality to state-of-the-art MCMC or variational inference methods that do not operate in context. The source code for this paper is available at https://github.com/ArikReuter/ ICL_for_Full_Bayesian_Inference
Additional Metrics?
[➜Log in]
Edit extra informations Login
Publication type Article: Conference contribution
Conference Title 42nd International Conference on Machine Learning, ICML 2025
Conference Date 13-19 July 2025
Conference Location Vancouver
Quellenangaben Volume: 267, Issue: , Pages: 51531-51582 Article Number: , Supplement: ,
Institute(s) Helmholtz Artifical Intelligence Cooperation Unit (HAICU)