Frontiers of Flows for Generative AI
About
A growing fraction of modern generative modeling can be viewed through a common lens: learning a flow that transports a simple distribution into the data distribution. This spans deterministic dynamics (including ODE-based flows) and stochastic dynamics (including diffusion and score models), with deep connections to ideas in optimal transport. This workshop brings together researchers working on the mathematical and algorithmic foundations of these models to explore the frontier of flows and generative modeling, and to identify key open problems that will drive the next generation of advances.
The workshop will consist of featured talks, lightning talks, and panel discussions on the frontiers of flows and generative AI. The goal is to surface emerging ideas and identify key open problems for the field.
Topics
- Flow-based generative models (flow matching, diffusion/score models, and related formulations)
- Sampling and learning algorithms
- One-step and few-step models; distillation and consistency-style training
- Optimal transport perspectives
- Connections with other generative paradigms
- Discrete flow and diffusion models
- Applications including vision, text, multimodal modeling, and scientific settings
Registration
Registration is now open!
CMU students, faculty, and staff: Free
Register (CMU) →
General public: $35
Register (General) →
Speakers
Location
Carnegie Museum of Art Theater
4400 Forbes Avenue, Pittsburgh, PA
Schedule
March 26 (Thursday): Workshop Day 1 (9:00am - 5:00pm)
| Time | Event |
|---|---|
| 9:00 - 9:30 | Welcome / Networking |
| 9:30 - 10:30 | Kickoff Panel: Mengyu Wang, Michael Albergo, Aditi Raghunathan Moderator: Sean Welleck |
| 10:30 - 11:15 | Featured Talk: Michael Albergo (Harvard) |
| 11:15 - 12:00 | Featured Talk: Mengyu Wang (Harvard) |
| 12:00 - 1:30 | Lunch (on your own) |
| 1:30 - 2:45 | Invited Lecture: Andrej Risteski (CMU) |
| 2:45 - 3:30 | Featured Talk: Nicholas Boffi (CMU) |
| 3:30 - 3:45 | Break / Networking |
| 3:45 - 4:30 | Featured Talk: Danilo Comminiello (Sapienza University of Rome) |
| 4:30 - 5:00 | Student Lightning Talks: 1. Shanda Li (CMU) - CodePDE: An Inference Framework for LLM-driven PDE Solver Generation 2. Garrett Mulcahy (UW) - Diffusion Approximation to Schrödinger Bridge |
March 27 (Friday): Workshop Day 2 (9:00am - 5:00pm)
| Time | Event |
|---|---|
| 9:00 - 10:00 | Student Lightning Talks: 1. Eric Frankel (UW) - S4S: Solving for a Diffusion Model Solver 2. Andre He (CMU) - Reasoning with Latent Tokens in Diffusion Language Models 3. Jerry Huang (CMU) - How to guide your flow: Steering flow maps for rapid test-time alignment |
| 10:00 - 10:30 | Break / Networking |
| 10:30 - 11:15 | Featured Talk: Michal Balcerak (University of Zurich) |
| 11:15 - 12:00 | Featured Talk: Aviral Kumar (CMU) |
| 12:00 - 1:15 | Lunch (on your own) |
| 1:15 - 2:45 | Invited Lecture: Sanjay Shakkottai (UT Austin) |
| 2:45 - 3:30 | Featured Talk: Yutong (Kelly) He (CMU) |
| 3:30 - 4:00 | Break / Networking |
| 4:00 - 5:00 | Closing Panel: Michal Balcerak, Yutong (Kelly) He, Nicholas Boffi Moderator: Zaid Harchaoui |
Organizers
Organizers
Student Organizers
Travel & Lodging
The workshop takes place at the Carnegie Museum of Art, located right next to the CMU campus. We recommend staying near CMU/University of Pittsburgh, though Downtown and other nearby areas also offer nice options.
CMU / University of Pittsburgh (Oakland, ~5-10min walk)
- The Oaklander Hotel
- Courtyard by Marriott Pittsburgh University Center
- Hilton Garden Inn Pittsburgh University Place
- Residence Inn Pittsburgh Oakland/University Place