LLaVA-OneVision-2: Towards Next-Generation Perceptual Intelligence

Benchmark Model Rank Results
video-question-answering-on-tempcompass-mcqLLaVA-OneVision-2-8B#3Accuracy: 74.5
visual-question-answering-on-v-benchLLaVA-OneVision-2 (8B)#7Accuracy: 85.9
visual-question-answering-on-docvqa-testLLaVA-OneVision-2-8B#10ANLS: 0.952
visual-question-answering-vqa-on-mmbench-enLLaVA-OneVision-2-8B#12Accuracy: 85.7
object-counting-on-countbenchLLaVA-OneVision-2-8B#17Accuracy: 89.0
video-question-answering-on-mlvuLLaVA-OneVision-2-8B#17Accuracy: 76.6
zero-shot-video-question-answer-on-video-mme-w-o-subsLLaVA-OneVision-2-8B#19Accuracy (%): 71.9
video-question-answering-on-lvbenchLLaVA-OneVision-2-8B#23Accuracy: 55.5
visual-question-answering-vqa-on-blinkLLaVA-OneVision-2-8B#23Accuracy: 63.5
video-question-answering-on-mmvuLLaVA-OneVision-2-8B#29Accuracy: 56.2
chart-question-answering-on-chartqaLLaVA-OneVision-2-8B#391:1 Accuracy: 85.9
video-question-answering-on-mvbenchLLaVA-OneVision-2-8B#46Avg.: 66.2
visual-question-answering-vqa-on-realworldqaLLaVA-OneVision-2-8B#53Accuracy: 69.7
optical-character-recognition-ocr-on-ocrbenchLLaVA-OneVision-2-8B#56Score: 782
vision-language-understanding-on-mmstarLLaVA-OneVision-2-8B#59Accuracy: 64.8