List your company
ResearchFormerly WizWand

Spatial Reasoning on SAT Real

75.3Accuracy (Pass@1)DCRL
29.571241.443153.31565.1869Apr 9, 2026Apr 18, 2026Apr 27, 2026May 6, 2026May 15, 2026May 24, 2026Jun 2, 2026
Updated 5mo ago

Evaluation results

MethodLinks
DCRLModel Category=Our Models2026.0675.3
World2VLM-GRPOSetting=SVC as WM2026.0472.67
Qwen3-VL-8BModel Category=Our Models2026.0670
World2VLM-GRPOSetting=HY-WorldPlay as WM2026.0469.33
GPT-4o (CoT)Reasoning Strategy=CoT2026.0468.67
World2VLM-SFTSetting=HY-WorldPlay as WM2026.0468.66
GRPO-TBackbone=Qwen2.5-VL-7B-Instruct, Model Scale=7B, Reasoning Strategy=RL (Task Reward Only)2026.0467
FGRPOBackbone=Qwen2.5-VL-7B-Instruct, Model Scale=7B, Reasoning Strategy=Faithful GRPO2026.0465.66
VL-RethinkerReasoning Strategy=MRM baseline2026.0465
GPT-5-nano (CoT)Reasoning Strategy=CoT2026.0464
World2VLM-SFTSetting=SVC as WM2026.0464
Non-reasoningBackbone=Qwen2.5-VL-7B-Instruct, Model Scale=7B, Reasoning Strategy=Non-reasoning2026.0463.11
TreeVGRReasoning Strategy=MRM baseline2026.0461
CoT promptingBackbone=Qwen2.5-VL-7B-Instruct, Model Scale=7B, Reasoning Strategy=CoT prompting2026.0459.22
Non-reasoningBackbone=Qwen2.5-VL-3B-Instruct, Model Scale=3B, Reasoning Strategy=Non-reasoning2026.0459
GRPO-TBackbone=Qwen2.5-VL-3B-Instruct, Model Scale=3B, Reasoning Strategy=RL (Task Reward Only)2026.0458.67
FGRPOBackbone=Qwen2.5-VL-3B-Instruct, Model Scale=3B, Reasoning Strategy=Faithful GRPO2026.0458.6
Vision-R1Reasoning Strategy=MRM baseline2026.0458.45
ViGoRL-SpatialReasoning Strategy=MRM baseline2026.0458.44
GPT-4oModel Category=Proprietary Model2026.0657.5
Qwen2.5-VL-7BModel Category=Open-Weight Multi Image Models2026.0656.33
CoT promptingBackbone=Qwen2.5-VL-3B-Instruct, Model Scale=3B, Reasoning Strategy=CoT prompting2026.0455
R1-OneVisionReasoning Strategy=MRM baseline2026.0451.5
Qwen2.5-VL-7BSetting=Base VLM2026.0444.67
Qwen2.5-VL-7B + WMSetting=MindJourney-style[55]2026.0431.33