List your company

Open-ended Video Question Answering on NExT-QA (test)

33.9AccuracyFrameOracle
16.01220.65625.329.944Oct 4, 2025

Evaluation results

MethodLinks
FrameOracleBackbone=LLaVA-Video-7B, Frames=64→13.92025.1033.9
FrameOracleBackbone=Qwen3-VL-8B, Frames=128→27.82025.1033.8
FrameOracleBackbone=VideoLLaMA3-7B, Frames=64→13.92025.1033.6
FrameOracleBackbone=LLaVA-Video-7B, Frames=16→10.42025.1033
FrameOracleBackbone=VideoLLaMA3-7B, Frames=16→10.42025.1032.9
LLaVA-Video-7BFrames=162025.1032.4
VideoLLaMA3-7BFrames=162025.1032.3
FrameOracleBackbone=Qwen3-VL-8B, Frames=32→20.92025.1032.3
FrameOracleBackbone=Qwen2.5-VL-3B, Frames=128→27.82025.1031.7
Qwen3-VL-8BFrames=322025.1031.1
FrameOracleBackbone=Qwen2.5-VL-3B, Frames=32→20.92025.1030.5
Qwen2.5-VL-3BFrames=322025.1029.6
FrameOracleBackbone=LLaVA-OneVision-7B, Frames=64→13.92025.1019
FrameOracleBackbone=LLaVA-OneVision-7B, Frames=16→10.42025.1017.8
LLaVA-OneVision-7BFrames=162025.1016.7