HKUSTGZ

Unified Diffusion VLA: Vision-Language-Action Model via Joint Discrete Denoising Diffusion Process

Less is More: Improving LLM Reasoning with Minimal Test-Time Intervention