This talk explores the paradox of why diffusion models generalize rather than memorize. It introduces an analytically tractable, wavelet-based parameterization of the score function, allowing for an interpretable analysis of how architectural biases (like locality) and data statistics interact to influence denoising performance and generalization.
Greg Kamradt of the ARC Prize Foundation explains how the ARC-AGI benchmark is shifting the focus of AI evaluation from memorization to true intelligence, defined as the ability to generalize and learn new skills efficiently. He discusses the history of ARC-AGI, how it revealed the limits of early LLMs and highlighted the recent "reasoning breakthrough," and details the upcoming interactive ARC-AGI v3, which will measure AI performance against a human baseline with zero instructions.
Quan Vuong and Jost Tobias Springenberg from Physical Intelligence (PI) discuss their mission to create a universal model for controlling any robot. They detail their approach, which centers on Vision-Language-Action (VLA) models, a purpose-built data engine for scaled data collection, and the evolution of their models toward open-world generalization.