AIM: Adaptive Inference of Multi-Modal LLMs via Token Merging and Pruning Paper • 2412.03248 • Published Dec 4, 2024 • 26
