Ovis2.5 Technical Report

Benchmark Model Rank Results
chart-question-answering-on-chartqaOvis2.5-9B#11:1 Accuracy: 92.9
visual-question-answering-vqa-on-textvqa-valOvis2.5-9B#1Accuracy: 91.2
visual-question-answering-vqa-on-mmt-benchOvis2.5-9B#3Accuracy: 69.3
optical-character-recognition-ocr-on-ocrbench-v2Ovis2.5-9B#4English Score: 63.4Chinese Score: 58.0
video-question-answering-on-tempcompass-overallOvis2.5-9B#8Accuracy: 72.6
visual-question-answering-vqa-on-hallusionbench-average-accuracyOvis2.5-9B#8Accuracy: 65.1
visual-grounding-on-refcoco-refcoco-refcocog-overallOvis2.5-9B#10Accuracy: 90.1
visual-question-answering-vqa-on-blinkOvis2.5-9B#12Accuracy: 67.3
vision-language-understanding-on-mmbench-v1-1Ovis2.5-9B#13Accuracy: 84.9
optical-character-recognition-ocr-on-ocrbenchOvis2.5-9B#15Score: 879
multimodal-reasoning-on-mathvistaOvis2.5-9B#18Accuracy: 83.4
video-question-answering-on-video-mme-with-subtitlesOvis2.5-9B (w/ subtitles)#18Accuracy: 72.8
video-question-answering-on-mlvuOvis2.5-9B#19Accuracy: 76.4
multimodal-reasoning-on-mathvisionOvis2.5-9B#28Accuracy: 53.9
multimodal-reasoning-on-mmmuOvis2.5-9B#28Accuracy: 71.2
video-question-answering-on-mvbenchOvis2.5-9B#32Avg.: 69.7
vision-language-understanding-on-mmstarOvis2.5-9B#32Accuracy: 72.4
multimodal-reasoning-on-mmmu-proOvis2.5-9B#37Accuracy: 54.4