Zooming without Zooming v2: From Seeing Details to Connecting Evidence for Generalizable Multimodal Perception

Benchmark Model Rank Results
computer-use-agents-on-ui-v-ui-visionZwZ2-27B#2Accuracy: 66.66
visual-question-answering-on-v-benchZwZ2-27B#2Accuracy: 91.62
visual-question-answering-vqa-on-erqaZwZ2-27B#2Accuracy: 61.75
spatial-reasoning-on-mmsi-benchZwZ2-27B#4Accuracy: 37.1
computer-use-agents-on-screenspot-proZwZ2-27B#6Accuracy: 70.59
multimodal-reasoning-on-babyvisionZwZ2-27B#10Accuracy: 28.35
vision-language-understanding-on-mmstarZwZ2-27B#13Accuracy: 79.07