REPA Works Until It Doesn't: Early-Stopped, Holistic Alignment Supercharges Diffusion Training Paper • 2505.16792 • Published May 22
Drag-and-Drop LLMs: Zero-Shot Prompt-to-Weights Paper • 2506.16406 • Published 9 days ago • 108
DSP: Dynamic Sequence Parallelism for Multi-Dimensional Transformers Paper • 2403.10266 • Published Mar 15, 2024
Real-Time Video Generation with Pyramid Attention Broadcast Paper • 2408.12588 • Published Aug 22, 2024 • 17
HeteGen: Heterogeneous Parallel Inference for Large Language Models on Resource-Constrained Devices Paper • 2403.01164 • Published Mar 2, 2024