Paper summary
Paper summary: Thinking with Video: Video Generation as a Promising Multimodal Reasoning Paradigm
"Thinking with Video" uses a video generation model (Sora-2) as a reasoner, drawing its solution as video frames; on a new benchmark it rivals strong VLMs on visual puzzles and reaches 92% on MATH.
- A video generation model can solve problems by generating a video of the solution being worked out.
- The authors could only evaluate Sora-2 from outside, because it isn't open-source.