6 papers
HEART: Exploiting Head Heterogeneity in Sparse Attention for Video Diffusion
Xuzhe Zheng, Yuexiao Ma, Jing Xu +3
Sparse attention accelerates video diffusion by allowing each attention head to focus on only a small subset of interactions. Existing methods already construct head-specific spars…
Motion-Aware Caching for Efficient Autoregressive Video Generation
Jing Xu, Yuexiao Ma, Xuzhe Zheng +7
Autoregressive video generation paradigms offer theoretical promise for long video synthesis, yet their practical deployment is hindered by the computational burden of sequential i…
Event-Anchored Frame Selection for Effective Long-Video Understanding
Wang Chen, Yongdong Luo, Yuhui Zeng +5
Massive frame redundancy and limited context window make efficient frame selection crucial for long-video understanding with large vision-language models (LVLMs). Prevailing approa…
Flow caching for autoregressive video generation
Yuexiao Ma, Xuzhe Zheng, Jing Xu +9
Autoregressive models, often built on Transformer architectures, represent a powerful paradigm for generating ultra-long videos by synthesizing content in sequential chunks. Howeve…
Video-RAG: Visually-aligned Retrieval-Augmented Long Video Comprehension
Yongdong Luo, Xiawu Zheng, Guilin Li +8
Existing large video-language models (LVLMs) struggle to comprehend long videos correctly due to limited context. To address this problem, fine-tuning long-context LVLMs and employ…
Polybasic Speculative Decoding Through a Theoretical Perspective
Ruilin Wang, Huixia Li, Yuexiao Ma +4
Inference latency stands as a critical bottleneck in the large-scale deployment of Large Language Models (LLMs). Speculative decoding methods have recently shown promise in acceler…