Generative Modeling with Bayesian Sample Inference

Name: Generative Modeling with Bayesian Sample Inference
Uploaded: 2026-08-20T13:33:31.563163+00:00
Duration: 508 s
Description: The episode explores the paper "Generative Modeling with Bayesian Sample Inference," a new framework that reframes data generation. Instead of adding noise, it treats sample creation as a process of refining belief. The discussion covers how this method achieves superior sample quality compared to existing models (BFNs and VDM) on benchmarks like ImageNet.

The episode explores the paper "Generative Modeling with Bayesian Sample Inference," a new framework that reframes data generation. Instead of adding noise, it treats sample creation as a process of refining belief. The discussion covers how this method achieves superior sample quality compared to existing models (BFNs and VDM) on benchmarks like ImageNet.

Open the full episode