List your company
ResearchFormerly WizWand

Video Question Answering on Video-MME Medium Duration 1.0

58.1Accuracy (No Subtitles)VILA-1.5
35.94841.69947.4553.201Nov 4, 2024Nov 6, 2024Nov 9, 2024Nov 11, 2024Nov 14, 2024Nov 16, 2024Nov 19, 2024
Updated 8mo ago

Evaluation results

MethodLinks
VILA-1.5LLM Size=34B2024.1158.157.4
LLaVA-NeXT-VideoLLM Params=34B2024.1150.152.2
LLaVA-NeXTLLM Size=34B, DPO tuning=true2024.1150.152.2
Qwen-VL-Max2024.1149.248.9
InternVL-V1.5LLM Params=20B2024.1146.449.1
PPLLAVALLM Params=7B, Multi-image training=true2024.1145.650.4
PPLLAVALLM Params=7B, Multi-image training=false2024.1143.948.6
DynFocusInput=16 f, LLM Size=7B, Resolution=224x224, L=25, K/L=0.82024.1143.746
Chat-UniVi-V1.5LLM Params=7B2024.1140.344.6
LLaVA-NeXT-VideoLLM Params=7B2024.1140.344.3
Chat-UniViLLM Size=7B2024.1140.344.6
Qwen-VL-ChatLLM Params=7B2024.1138.740.4
Video-LLaVALLM Params=7B2024.113840.7
Video-LLaVAInput=8 f, LLM Size=7B2024.113840.7
VideoChat2-MistralLLM Params=7B2024.113739.4
VideoChat2Input=16 f, LLM Size=7B2024.113739.4
ST-LLMLLM Params=7B2024.1136.841.4
ST-LLMInput=16 f, LLM Size=7B2024.1136.841.4