PaliGemma: A versatile 3B VLM for transfer

Benchmark Model Rank Results
visual-question-answering-on-msrvtt-qaPaliGemma#1Accuracy: 0.501
image-captioning-on-textcapsPaliGemma#2CIDEr: 153.9
visual-reasoning-on-nlvr2-testPaliGemma#2Accuracy: 90
visual-question-answering-on-ok-vqaPaliGemma#4Accuracy: 63.5
visual-question-answering-vqa-on-mmvpPaliGemma#4Paired Accuracy: 47.3
visual-question-answering-on-docvqa-testPaliGemma#60ANLS: 0.848