PixelPrune: Pixel-Level Adaptive Visual Token Reduction via Predictive Coding
Open paper
Benchmark
Model
Rank
Results
computer-use-agents-on-screenspot-v2
Qwen3-VL (2B) with PixelPrune (+KD)
#31
Accuracy: 90.6
computer-use-agents-on-screenspot-pro
Qwen3-VL (2B) with PixelPrune (+KD)
#33
Accuracy: 47.6
Rank counts only results with a code link.