Zehuan Wu

InterSketch shows that interleaving visual sketches with textual reasoning, guided by self-correction and stepwise rewards, unlocks surprisingly strong long-horizon visual reasoning, even surpassing Gemini-3-Pro.

Generate safety-critical driving scenarios with full trajectory control, even *beyond* your training data, using RL to fine-tune a video diffusion model.

Edit the appearance of driving scenes and generate consistent free-viewpoint images along novel trajectories, opening new possibilities for autonomous driving simulation.