2025
A-VL: Adaptive Attention for Large Vision-Language Models
AAAI 2025technical
The Large Vision-Language Model (LVLM) integrates computer vision and natural language processing techniques, offering substantial application potential. However, these models demand extensive resources during inference. Adaptive attention techniques can dynamically reduce computational redundancy a…