AINeutralarXiv – CS AI · 3h ago6/10
🧠
AdaMerge: Salience-Aware Adaptive Token Merging for Training-Free Acceleration of Vision Transformers
AdaMerge introduces a training-free method to accelerate Vision Transformers by improving token merging through salience-aware mechanisms and adaptive layer-wise compression. The approach outperforms existing token reduction methods across all computational efficiency benchmarks, maintaining superior accuracy-to-FLOPs ratios on ImageNet-1k evaluations.