MHLA: Restoring Expressivity of Linear Attention via Token-Level Multi-Head Paper • 2601.07832 • Published Jan 12 • 52
UniLumos: Fast and Unified Image and Video Relighting with Physics-Plausible Feedback Paper • 2511.01678 • Published Nov 3, 2025 • 38
Stencil: Subject-Driven Generation with Context Guidance Paper • 2509.17120 • Published Sep 21, 2025 • 6
VChain: Chain-of-Visual-Thought for Reasoning in Video Generation Paper • 2510.05094 • Published Oct 6, 2025 • 38