List your company
ResearchFormerly WizWand

Video Question Answering on ActivityNet-QA LLaVA-Hound in-domain (test)

68.5AccuracyMM1.5-Video
32.82842.08951.3560.611Sep 30, 2024
Updated 8mo ago

Evaluation results

MethodLinks
MM1.5-VideoModel Scale=7B, Protocol=SFT2024.0968.5
MM1.5-VideoModel Scale=3B, Protocol=SFT2024.0967.8
MM1.5-VideoModel Scale=1B, Protocol=SFT2024.0965.7
LLAVA-HOUND-SFTModel Scale=7B2024.0962.8
MM1.5-VideoModel Scale=7B, Protocol=Training-free2024.0952.8
MM1.5-VideoModel Scale=3B, Protocol=Training-free2024.0951.5
MM1.5-VideoModel Scale=1B, Protocol=Training-free2024.0949
Video-LLaVAModel Scale=7B2024.0941.4
Chat-UniViModel Scale=7B2024.0939.4
LLaMA-VIDModel Scale=7B2024.0936.5
Video-ChatGPTModel Scale=7B2024.0934.2