Probabilistic Inference for Controlling Diffusion Models @MicrosoftResearch
Probabilistic Inference for Controlling Diffusion Models  @MicrosoftResearch
Uploaded September 2026 | Updated November 2026, 3 hours ago
Diffusion models are probabilistic models, and controlling a diffusion model at inference time is a probabilistic inference problem. In this talk, we will take a tour through inference-time control for diffusion models and organize the landscape around two broad ideas: sequential control, where we guide and refine samples as generation proceeds, and parallel control, where multiple samples along denoising trajectories interact towards the target distribution. For each family, we will look at both approximate and exact methods: what they are doing probabilistically, where the approximations come from, and when exactness can be recovered. Together, these perspectives give us a simple roadmap for understanding the methods, their trade-offs, and when each style of control is most useful.

Speaker bio: Jiajun He (jiajunhe98.github.io/) is a PhD student in the Machine Learning Group at the University of Cambridge, and currently interning at Microsoft Research New England. He works broadly on probabilistic machine learning, with a particular interest in the interaction between path-space probabilistic methods and generative models. His research spans diffusion and flow-based models, nonequilibrium methods, sampling, and information theory.

Find seminar details and upcoming talks: microsoft.com/en-us/research/event/microsoft-research-new-england-generative-modeling-sampling-seminar
Probabilistic Inference for Controlling Diffusion ModelsMicrosoft Research Focus – Edition #97Aegon: Self-Auditable Key TransparencySample and Map from a Single Convex Potential: Generation using Conjugate Moment MeasuresIdeas: AI for materials discovery with Tian Xie and Ziheng LuAI, Radio Astronomy, and the Search for Life Beyond EarthAI that runs your repo in GitHub : New from Microsoft Research #ai #agenticai #github #workflowAI that runs your repo in GitHub : New from Microsoft Research #ai #agenticai #github #workflowTool-space Interference: An emerging problem for LLM agentsMicrosoft Research Forum | Season 2, Episode 4PhylaFlow: Hybrid flow matching in phylogenetic tree spaceAI Testing and Evaluation: Learnings from pharmaceuticals and medical devices
Microsoft Research |

Probabilistic Inference for Controlling Diffusion Models

SHARE TO X SHARE TO REDDIT SHARE TO FACEBOOK WALLPAPER