Qwen-Drive-1.0: An Initial Step towards a Vision-Language Foundation Model for Autonomous Driving

Benchmark Model Rank Results
visual-question-answering-vqa-on-realworldqaQwen-Drive-1.0-SFT#15Accuracy: 78.95
vision-language-understanding-on-mmstarQwen-Drive-1.0-SFT#20Accuracy: 75.87
multimodal-reasoning-on-mmmuQwen-Drive-1.0-SFT#21Accuracy: 72.67
navsim-on-opensceneQwen-Drive-1.0-RL#21PDMS: 90.7
optical-character-recognition-ocr-on-ocrbenchQwen-Drive-1.0-SFT#26Score: 864