List your company
ResearchFormerly WizWand

Video Multimodal Evaluation on VideoMME

61.48ScoreVideoRoPE
19.068830.079441.0952.1006Oct 27, 2025Nov 24, 2025Dec 22, 2025Jan 19, 2026Feb 16, 2026Mar 16, 2026Apr 14, 2026
Updated 6mo ago

Evaluation results

MethodLinks
VideoRoPEBackbone=Qwen3-VL-8B-Instruct2025.1061.48
MHRoPEBackbone=Qwen3-VL-8B-Instruct2025.1060.89
MRoPE-IBackbone=Qwen3-VL-8B-Instruct2025.1060.89
HoPEBackbone=Qwen3-VL-8B-Instruct2025.1060.56
Full ModelBackbone=LLaVA-OneVision, Sparsity Ratio=0%2026.0460.1
Vanilla RoPEBackbone=Qwen3-VL-8B-Instruct2025.1060.07
MRoPEBackbone=Qwen3-VL-8B-Instruct2025.1059.78
TopoVLMBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0458.2
LLM-StreamlineBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0455.6
SparseGPTBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0455.2
ECoFLaPBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0454.4
TAMPBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0454
LLM-PrunerBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0452.3
WandaBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0451.5
CircleRoPEBackbone=Qwen3-VL-8B-Instruct2025.1051.04
OWLBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0450.8
MagnitudeBackbone=LLaVA-OneVision, Sparsity Ratio=60%2026.0420.7