Xiaodong Yu

ImageDoctor: Diagnosing Text-to-Image Generation via Grounded Image Reasoning

KeyVID: Keyframe-Aware Video Diffusion for Audio-Synchronized Visual Animation

Latent Visual Reasoning

ZeroTuning: Unlocking the Initial Token's Power to Enhance Large Language Models Without Training

XModBench: Benchmarking Cross-Modal Capabilities and Consistency in Omni-Language Models

DRIFT: Directional Reasoning Injection for Fine-Tuning MLLMs

APRIL: Active Partial Rollouts in Reinforcement Learning to Tame Long-tail Generation

Where LLM Agents Fail And How They can Learn From Failures

Unleashing Hour-Scale Video Training for Long Video-Language Understanding

Model Tells Itself Where to Attend: Steerable Prompting for Reliable Reading Comprehension of LLM