Generative Multimodal Models are In-Context Learners

Benchmark Model Rank Results
personalized-image-generation-on-dreambenchEmu2 SDXL v1.0#3Overall (CP * PF): 0.364Concept Preservation (CP): 0.528
visual-question-answering-on-mm-vetEmu2-Chat#52GPT-4 score: 48.5Params: 37B
visual-question-answering-on-mm-vet-v2Emu2-Chat#12GPT-4 score: 38.0±0.1