List your company

Video Question Answering on Perception (test)

75.6Test AccuracyVideoChat-Flash
-2.3282417.9031338.134558.36587Apr 1, 2024Jul 17, 2024Nov 2, 2024Feb 18, 2025Jun 5, 2025Sep 21, 2025Jan 7, 2026

Evaluation results

MethodLinks
VideoChat-FlashSize=7B, #Tokens=162025.0175.6————
InternVideo2.5 (InternVL2.5+LRC)Size=7B, #Tokens=162025.0174.9————
Qwen2.5-VL-7B+TIR-FlowFrames=322026.0173.2————
VideoLLaMA3-7BFrames=1802026.0172.8————
LLaVA-Video-7B+TIR-FlowFrames=322026.0171.5————
Qwen2.5-VL-7BFrames=322026.0169.1————
BIMBA-LLaVALLM=Qwen2-7B, Frames=1282025.0368.51————
InternVL2.5Size=7B, #Tokens=2562025.0168.2————
QwenVL2Size=72B2025.0168————
LLaVA-VideoLLM=Qwen2-7B, Frames=642025.0367.9————
LLaVA-Video-7BFrames=322026.0167.9————
LLaVA-OneVisionSize=72B, #Tokens=1962025.0166.9————
Qwen2-VL-7B-InstructFrames=322026.0163.7————
InternVideo2-HDSize=7B, #Tokens=722025.0163.4————
QwenVL2Size=7B2025.0162.3————
Qwen2 VL (7B)# Frames=2 fps (max 768)2024.1062.3————
VideoLLaMA2Size=72B, #Tokens=722025.0157.5————
LLaVA-OneVisionSize=7B, #Tokens=1962025.0157.1————
LLaVA-OneVisionLLM=Qwen2-7B, Frames=322025.0357.1————
Baichuan-omni (7B)# Frames=1 fps (max 48)2024.1056.8————
BIMBA-LLAMALLM=LLaMA3.2-8B, Frames=642025.0356.5————
VITA (8x7B)# Frames=1 fps (max 32)2024.1056.2————
VideoChat-TPOSize=7B, #Tokens=642025.0154.7————
VILA-40BFrames=2562026.0154————
LLaMA-3.2 (Video)LLM=LLaMA3.2-8B, Frames=642025.0353.16————
PLLAVA (LLaMA-3.2)LLM=LLaMA3.2-8B, Frames=642025.0353.13————
BIMBA-LLaVALLM=Vicuna-7B, Frames=642025.0352.61————
VideoLLaMA2Size=7B, #Tokens=722025.0151.4————
VideoLLaMA 2 (7B)# Frames=162024.1051.4————
Video-LLaMA2LLM=Qwen2-7B, Frames=322025.0351.4————
TraveLERTraining Protocol=Zero-shot2024.0450.2————
LLaVA-NeXT-VideoSize=7B, #Tokens=1442025.0148.8————
LLaVA-NeXT-Video (7B)# Frames=322024.1048.8————
LLaVA-NeXT-VideoLLM=Vicuna-7B, Frames=322025.0348.8————
PLLAVALLM=Vicuna-7B, Frames=642025.0348.55————
MC-ViT-LTraining Protocol=Uses fine-tuned components2024.0448.1————
VideoChat2 (7B)# Frames=162024.1047.3————
VideoChat2LLM=Vicuna-7B, Frames=162025.0347.3————
MC-ViT-BTraining Protocol=Uses fine-tuned components2024.0447————
SeViLATraining Protocol=Uses fine-tuned components2024.0446.2————
LLaVA-NeXT (Video)LLM=Vicuna-7B, Frames=642025.0346.13————
Long ViViTTraining Protocol=Zero-shot2024.0445.7————
BOLDModel=SeViLA, k=0.52024.1045.3245.185.182.951.3
Weighted_BOLDModel=SeViLA, k=0.52024.1045.3145.24.562.831.08
LLaMA-VIDSize=7B, #Tokens=22025.0144.6————
LLAMA-VIDLLM=Vicuna-7B, Frames=1fps2025.0344.6————
Video-LLaVA (7B)# Frames=82024.1044.3————
Video-LLaVALLM=Vicuna-7B, Frames=82025.0344.3————
Flamingo-3BTraining Protocol=Zero-shot2024.0443.6————
Weighted_BOLDModel=Video-LLaMA, k=0.52024.1042.0639.3621.89.1113.02
Weighted_BOLDModel=Video-LLaVA, k=0.52024.1041.9539.4619.7310.243.4
BOLDModel=Video-LLaMA, k=0.52024.1041.938.6824.059.8714.45
BOLDModel=Video-LLaVA, k=0.52024.1041.6238.6921.7510.923.78
AnyGPT (8B)# Frames=482024.1029.1————
V-JEPA 2Params Enc / LLM=1B/8B, Evaluation Protocol=SFT2025.060.84————
PLM 8BParams Enc / LLM=1B/8B, Evaluation Protocol=zero-shot2025.060.827————
Qwen2.5VLParams Enc / LLM=1B/7B, Evaluation Protocol=zero-shot2025.060.705————
InternVL-2.5Params Enc / LLM=300M/7B, Evaluation Protocol=zero-shot2025.060.689————
Qwen2VLParams Enc / LLM=675M/7B, Evaluation Protocol=zero-shot2025.060.669————