List your company
ResearchFormerly WizWand

Video Question Answering on TGIF-QA LLaVA-Hound out-of-domain (test)

65.5AccuracyMM1.5-Video
25.66836.00946.3556.691Sep 30, 2024
Updated 8mo ago

Evaluation results

MethodLinks
MM1.5-VideoModel Scale=7B, Protocol=SFT2024.0965.5
MM1.5-VideoModel Scale=3B, Protocol=SFT2024.0964.9
MM1.5-VideoModel Scale=1B, Protocol=SFT2024.0964
LLAVA-HOUND-SFTModel Scale=7B2024.0961.1
MM1.5-VideoModel Scale=7B, Protocol=Training-free2024.0949.8
MM1.5-VideoModel Scale=3B, Protocol=Training-free2024.0949.2
MM1.5-VideoModel Scale=1B, Protocol=Training-free2024.0947.6
Chat-UniViModel Scale=7B2024.0933.2
Video-LLaVAModel Scale=7B2024.0933
Video-ChatGPTModel Scale=7B2024.0931.4
LLaMA-VIDModel Scale=7B2024.0927.2