CogAgent: A Visual Language Model for GUI Agents

Benchmark Model Rank Results
computer-use-agents-on-mind2web-cross-domainCogAgent-18B#6Step Success Rate: 59.4
computer-use-agents-on-mind2web-cross-taskCogAgent-18B#6Step Success Rate: 62.3
computer-use-agents-on-mind2web-cross-websiteCogAgent-18B#6Step Success Rate: 54.0
visual-question-answering-on-mm-vet-v2CogAgent-Chat#13GPT-4 score: 34.7±0.2
natural-language-visual-grounding-on-screenspotCogAgent#30Accuracy (%): 47.4
visual-question-answering-on-mm-vetCogAgent#60GPT-4 score: 52.8Params: 18B