AI Search reviews research into transferring motion between video characters and building controllable representations of moving objects. It also describes a robot world model that uses motion structure to generalize across different robot bodiesRobotics develops machines that sense their surroundings and perform physical actions through coordinated hardware and software..
The roundup discusses real-time translation, text diffusion and newer open-weight language models. It explains sparse attention as a way to select relevant parts of long context before more expensive processing, while attributing performance comparisons to the presented reports.
The presenter examines agent evaluationsEvaluation is the systematic assessment of a system's behaviour or results against explicit criteria. and research workflows, then discusses reported restrictions on Claude Fable access and concerns about safety behavior. Other covered work includes 3D human reconstruction, scene generation, speech and longer video generationA video generation model creates or transforms moving images from text, images, video, audio, or structured controls..
Watch on YouTube




