OneThinker: All-in-one Reasoning Model for Image and Video

Benchmark Model Rank Results
video-question-answering-on-videommmuOneThinker-8B#4Accuracy: 66.2
visual-question-answering-vqa-on-mmt-benchOneThinker-8B#5Accuracy: 67.8
referring-video-object-segmentation-on-mevisOneThinker-8B#6J&F: 52.7
referring-video-object-segmentation-on-reasonvosOneThinker-8B#8J&F: 54.9
mathematical-reasoning-on-mathverseOneThinker-8B#9Accuracy: 64.3
video-question-answering-on-videoholmesOneThinker-8B#9Accuracy: 48.7
video-question-answering-on-video-mmmuOneThinker-8B#19Accuracy: 66.2
multimodal-reasoning-on-mmmuOneThinker-8B#30Accuracy: 70.6
multimodal-reasoning-on-mathvistaOneThinker-8B#37Accuracy: 77.6
vision-language-understanding-on-mmstarOneThinker-8B#39Accuracy: 70.6