Chengru Song

HeadCast accelerates autoregressive video generation by up to 1.95x without retraining, preserving long-range consistency while optimizing attention head usage.

JAGG slashes the backward pass cost in GRPO training of diffusion models by nearly 50%, enabling efficient high-resolution text-to-image generation.

Achieving up to 1.90X speedup in video generation without sacrificing fidelity, ScalingAttention redefines efficiency in Diffusion Transformers.

Achieving lossless processing of 256K contexts, Keye-VL-2.0 transforms how we approach long-video understanding and agentic intelligence.

Surprisingly, the "think before answer" paradigm fails to enhance generative recommendation models, prompting a novel approach that redefines how reasoning is integrated into these systems.

Sub-linear attention is now possible without sacrificing complete long-range dependency retention, thanks to learnable summary tokens that compress context.

Stop handcuffing student diffusion models to their teachers: framing distribution matching as a reward unlocks more stable and performant distillation via RL techniques.