Leveraging Multi-Modal Representations — S. Jegelka, TUM / MIT | HAICON26

Speaker: S. Jegelka Affiliation: Technical University of Munich (TUM) / Massachusetts Institute of Technology (MIT) Date: June 10, 2026 Time: 11:30am–12:00pm (Invited Talk) Session: AI Research across Bavaria (co-organized by BAIOSPHERE) Recorded at the Helmholtz AI Conference (HAICON26), June 9–11, 2026. More info: https://haicon.cc/conference-program/ About this talk: Leveraging Multi-Modal Representations S. Jegelka 1 1 Technical University of Munich (TUM), Germany / Massachusetts Institute of Technology (MIT), USA This invited talk explores methods for leveraging multi-modality: cross-modal guidance for diffusion models, and using unpaired multimodal data — viewing different modalities as projections of a shared underlying reality — to obtain practical and theoretical gains.