List your company
ResearchFormerly WizWand

Video reasoning on MLVU (test)

62.75AccuracyOmniJigsaw (CMM)
47.628451.554255.4859.4058Nov 25, 2025Dec 17, 2025Jan 9, 2026Jan 31, 2026Feb 23, 2026Mar 17, 2026Apr 9, 2026
Updated 5mo ago

Evaluation results

MethodLinks
OmniJigsaw (CMM)Inference Mode=w/o Audio2026.0462.75
OmniJigsaw (CMM)Inference Mode=w/ Audio2026.0462.35
OmniJigsaw (SMS)Inference Mode=w/ Audio2026.0461.75
VideoJigsawInference Mode=w/o Audio2026.0461.55
OmniJigsaw (SMS)Inference Mode=w/o Audio2026.0460.76
VideoJigsawInference Mode=w/ Audio2026.0460.16
OmniJigsaw (JMI)Inference Mode=w/o Audio2026.0459.76
OmniJigsaw (JMI)Inference Mode=w/ Audio2026.0458.76
Qwen3-Omni-30BInference Mode=w/o Audio2026.0458.76
Qwen3-Omni-30BInference Mode=w/ Audio2026.0457.97
Video-R1-7B + replayFrame Count=16-frame, Activation Replay=true2025.1157.4
Video-R1-7BFrame Count=16-frame, Activation Replay=false2025.1155.7
GPT-4oFrame Count=16-frame, Activation Replay=false2025.1154.2
Omni-R1Inference Mode=w/ Audio2026.0452.59
Video-R1Inference Mode=w/o Audio2026.0452.39
Omni-R1Inference Mode=w/o Audio2026.0451
HumanOmniV2Inference Mode=w/ Audio2026.0449
Qwen2.5-VL-7BFrame Count=16-frame, Activation Replay=false2025.1148.4
HumanOmniV2Inference Mode=w/o Audio2026.0448.21