An Image is Worth 1/2 Tokens After Layer 2: Plug-and-Play Inference Acceleration for Large Vision-Language Models
Research paper by Liang Chen and collaborators. Published 2024-03-11. Topics: Large Vision-Language Models, LLaVA-1.5, QwenVL-Chat, Video-LLaVA, attention computation.
Resource information
- Authors
- Liang Chen, Haozhe Zhao, Tianyu Liu, Shuai Bai, Junyang Lin, Chang Zhou, Baobao Chang
- Research topics
- Large Vision-Language Models, LLaVA-1.5, QwenVL-Chat, Video-LLaVA, attention computation, visual tokens, sparser approach, FastV, adaptive attention patterns, pruning, FLOPs, edge devices
- Publication date
- 2024-03-11
- Publication
- arXiv
- Last record update
- Nov 3, 2026