List your company
ResearchFormerly WizWand

Complex Reasoning on vcrbench

53.1AccuracyVideoLatent-7B (Ours)
23.56431.23238.946.568Mar 18, 2026Apr 3, 2026Apr 19, 2026May 5, 2026May 21, 2026Jun 6, 2026Jun 22, 2026
Updated 4mo ago

Evaluation results

MethodLinks
VideoLatent-7B (Ours)Frames=32, Architecture Category=Latent MLLM2026.0653.1
Open-o3-Video-7BFrames=32, Architecture Category=Standard MLLM2026.0652.5
VideoLatent-7B (Ours)Frames=64, Architecture Category=Latent MLLM2026.0652.5
Qwen2.5-VL-7BFrames=64, Architecture Category=Standard MLLM2026.0651.4
Video-R1-7BFrames=64, Architecture Category=Standard MLLM2026.0650.8
Mull-Token-7BFrames=32, Architecture Category=Latent MLLM2026.0650.6
Mull-Token-7BFrames=64, Architecture Category=Latent MLLM2026.0650.6
Qwen2.5-VL-7BFrames=32, Architecture Category=Standard MLLM2026.0650.4
VideoLatent-7B (Ours)Frames=16, Architecture Category=Latent MLLM2026.0649.6
Open-o3-Video-7BFrames=64, Architecture Category=Standard MLLM2026.0649.2
LVR-7BFrames=32, Architecture Category=Latent MLLM2026.0649
Mull-Token-7BFrames=16, Architecture Category=Latent MLLM2026.0648.6
LVR-7BFrames=64, Architecture Category=Latent MLLM2026.0648
Video-R1-7BFrames=32, Architecture Category=Standard MLLM2026.0646.9
Qwen2.5-VL-7BFrames=16, Architecture Category=Standard MLLM2026.0646.7
Video-R1-7BFrames=16, Architecture Category=Standard MLLM2026.0646.3
LVR-7BFrames=16, Architecture Category=Latent MLLM2026.0645.3
Qwen3-VL-8B + SynRLModel Scale=8B, SynRL Training=true2026.0335.6
Qwen3-VL-8BModel Scale=8B, SynRL Training=false2026.0334.2
Qwen3-VL-4B + SynRLModel Scale=4B, SynRL Training=true2026.0331.5
Qwen3-VL-4BModel Scale=4B, SynRL Training=false2026.0330.1
Monet-7BFrames=32, Architecture Category=Latent MLLM2026.0625.5
Monet-7BFrames=64, Architecture Category=Latent MLLM2026.0625.1
Monet-7BFrames=16, Architecture Category=Latent MLLM2026.0624.7