Covering Human Action Space for Computer Use: Data Synthesis and Benchmark
Open paper
Benchmark
Model
Rank
Results
computer-use-agents-on-ui-v-ui-vision
Phi-Ground-Any-4B + APP Data
#7
Accuracy: 29.7
computer-use-agents-on-screenspot-pro
Phi-Ground-Any-4B + APP Data
#40
Accuracy: 41.5
Rank counts only results with a code link.