Invited Talk -- Yiran Huang & Zeynep Akata (Technical University of Munich) -- Title: The Asymmetry of Adaptation: Reverse-Engineering Multimodal In-Context Learning
Zeynep Akata
2025 Invited Talk
in
Workshop: AI That Keeps Up: Workshop on Continual and Compatible Foundation Model Updates (CCFM)
in
Workshop: AI That Keeps Up: Workshop on Continual and Compatible Foundation Model Updates (CCFM)
Abstract
As foundation models face the pressure to rapidly adapt to dynamic global changes, In-Context Learning (ICL) offers a mechanism for adaptation without expensive parameter updates. However, the mechanisms governing how Large Language Models (LLMs) extend this capability to new, unseen modalities remain poorly understood. In this talk, I present a systematic reverse-engineering of multimodal ICL, utilizing controlled synthetic tasks to isolate the architectural and statistical drivers of adaptation.
Speaker
Zeynep Akata
Video
Chat is not available.
Successful Page Load