List your company
ResearchFormerly WizWand

Multimodal Reasoning on MMVP

76AccuracyQwen2.5-VL-VIF
22.64836.49950.3564.201Dec 13, 2025Jan 8, 2026Feb 3, 2026Mar 1, 2026Mar 27, 2026Apr 22, 2026May 18, 2026
Updated 5mo ago

Evaluation results

MethodLinks
Qwen2.5-VL-VIFParams=7B, Base model=Qwen-2.5-VL2026.0576——
Unsilencing Latent ReasoningBackbone=VLAA Thinking-7B2026.0573.67——
Qwen2.5-VL-SFTParams=7B, Base model=Qwen-2.5-VL2026.0573——
DMLRBackbone=R1 OneVision-7B2026.0571.9——
Unsilencing Latent ReasoningBackbone=R1 OneVision-7B2026.0571——
Qwen2.5-VLParams=7B, Base model=Qwen-2.5-VL2026.0570.33——
ICoTBackbone=R1 OneVision-7B2026.0569.6——
DMLRBackbone=VLAA Thinking-7B2026.0569.4——
CCoTBackbone=R1 OneVision-7B2026.0568.9——
VanillaBackbone=VLAA Thinking-7B2026.0568.3——
ICoTBackbone=VLAA Thinking-7B2026.0568.3——
CCoTBackbone=VLAA Thinking-7B2026.0568——
MCoTBackbone=R1 OneVision-7B2026.0568——
MCoTBackbone=VLAA Thinking-7B2026.0567.2——
VanillaBackbone=R1 OneVision-7B2026.0567——
LLaVA-1.5-VIFParams=13B, Base model=LLaVA-1.52026.0566.67——
LLaVA-1.5-SFTParams=13B, Base model=LLaVA-1.52026.0565——
LLaVA-1.5Params=13B, Base model=LLaVA-1.52026.0563.33——
LLaVA-1.5-VIFParams=7B, Base model=LLaVA-1.52026.0553.33——
Cambrian-1Params=8B2026.0551.3——
GPT-4V-1106Params=-2026.0550——
LLaVA-1.5-SFTParams=7B, Base model=LLaVA-1.52026.0547.33——
ThinkLite-VL (+ Ours)Model=ThinkLite-VL, Method=Self-Reflection2025.1239.71.7—
RossParams=7B2026.0539.3——
ThinkLite-VL (Vanilla)Model=ThinkLite-VL, Method=Vanilla2025.12358.4—
LLaVA-1.5Params=7B, Base model=LLaVA-1.52026.0524.7——
DAM-3Bmask=12025.10——60.7
GAR-8Bmask=12025.10——78
GAR-8Bmask=1, Training Data=w/ 600K General Data2025.10——79.7
PAM-3Bmask=12025.10——4.3
PerceptionLM-8Bmask=12025.10——76