List your company
ResearchFormerly WizWand

Spatial Reasoning on MindCube (tiny)

94.5AccuracyHuman
17.976837.843457.7177.5766Nov 10, 2025Dec 13, 2025Jan 16, 2026Feb 19, 2026Mar 25, 2026Apr 28, 2026Jun 1, 2026
Updated 3mo ago

Evaluation results

MethodLinks
HumanCategory=Reference2026.0594.5
ProSRDescription=QWEN3-VL-8B-THINKING + Ours2026.0588.6
OursFeatures=Active, Cog. Map2026.0685
GeoThinkerCategory=Open-source Spatial Intelligence Models2026.0583
3DThinkerQwen2.5-VL-3BFeatures=Passive, 3D Rec.2026.0681.8
OursFeatures=Active, Cog. Map2026.0681
OursFeatures=Active, Cog. Map2026.0680.5
MindCubeQwen2.5-VL-3BFeatures=Passive, Cog. Map2026.0679.2
QWEN3-VL-8B-THINKING + GRPOMethod Variant=Vanilla GRPO2026.0575.8
3DThinkerQwen2.5-VL-3BFeatures=Passive, 3D Rec.2026.0675.2
SenseNova-SI-1.1-Qwen3-VL-8BCategory=Open-source Spatial Intelligence Models2026.0573.8
Gemini 3 ProCategory=Proprietary Models2026.0570.9
SenseNova-U1Parameters=30BA3B, Internal Reasoning (Think mode)=true2026.0570.86
MindCubeQwen2.5-VL-3BFeatures=Passive, Cog. Map2026.0670.7
QWEN3-VL-8B-THINKING + SFTMethod Variant=SFT Only2026.0568.4
Qwen3.5Parameters=35BA3B, Internal Reasoning (Think mode)=false2026.0563.46
SenseNova-U1Parameters=8B, Internal Reasoning (Think mode)=true2026.0562.01
GPT-3.2Category=Proprietary Models2026.0560.4
Gemini 2.5 ProCategory=Proprietary Models2026.0557.6
Qwen3.5Parameters=9B, Internal Reasoning (Think mode)=false2026.0557.59
3DThinkerQwen2.5-VL-3BFeatures=Passive, 3D Rec.2026.0655.5
Gemma4Parameters=26BA4B, Internal Reasoning (Think mode)=false2026.0548.84
Seed 1.6Category=Proprietary Models2026.0548.8
Claude-4-Sonnet-20250514Features=Passive2026.0648.42
MindCubeQwen2.5-VL-3BFeatures=Passive, Cog. Map2026.0648
SpatialThinker-30BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1145.4
SpatialThinker-7BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1145.1
Claude-4-Sonnet-0514Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1144.8
Claude-4-Sonnet-20250514Features=Passive2026.0644.75
Claude-4-Sonnet-20250514Features=Passive2026.0644.21
Qwen2.5-VL-7B + Vanilla GRPOEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1144
SpatialLadder-3BCategory=Open-source Spatial Intelligence Models2026.0543.5
Qwen3VLParameters=8B, Internal Reasoning (Think mode)=true2026.0543.17
GPT-5-0807Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1142.5
Qwen2.5-VL-7B + SFTEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1142.2
Qwen3VLParameters=30BA3B, Internal Reasoning (Think mode)=true2026.0540.86
Qwen3-VL-8B-ThinkingCategory=Open-source General Models2026.0540.6
SpatialThinker-3BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1140.5
InternVL3.5-8BCategory=Open-source General Models2026.0540.2
GPT-4oFeatures=Passive2026.0640.17
VST-7B-SFTCategory=Open-source Spatial Intelligence Models2026.0539.7
Argos2025.1239.6
Qwen3-VL-30BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1139
GPT-4oFeatures=Passive2026.0638.81
Qwen2.5-VL-3B + Vanilla GRPOEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1138.8
Qwen2.5-VL-7B-InstructFeatures=Passive2026.0638.76
Spatial-MLLMFeatures=Passive, 2D+3D2026.0638.39
Random (frequency)Features=–2026.0638.3
Space-QwenFeatures=Passive2026.0638.02
SpaceR-SFT-7BCategory=Open-source Spatial Intelligence Models2026.0538
Cambrian-S-7BCategory=Open-source Spatial Intelligence Models2026.0537.9
Qwen2.5-VL-3B-InstructFeatures=Passive2026.0637.37
VLAA-Thinker-7BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1136.7
Qwen2.5-VL-3BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1136.4
Random (chance)Features=–2026.0636.36
Qwen2.5-VL-7B-InstructCategory=Open-source General Models2026.0536
SpaceThinkerEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1136
SATORI-R1Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1136
Visionary-R1Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1136
Qwen2.5-VL-3B + SFTEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1135.9
Qwen2.5-VL-7BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1135.3
Qwen2.5VL 7B2025.1234.9
GPT-4o-0513Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1134.9
BAGEL-7B-MoTCategory=Open-source General Models2026.0534.7
SpatialReasonerEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1134.6
Video-R1Training Protocol=SFT2025.1234.2
Space-QwenFeatures=Passive2026.0633.71
SpaceOmEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1133.5
Space-QwenFeatures=Passive2026.0633.28
Qwen2.5-VL-3B-InstructFeatures=Passive2026.0633.26
Qwen2.5-VL-3B-InstructFeatures=Passive2026.0633.21
Qwen2.5VL 7BPrompting Strategy=CoT2025.1233.1
Random (frequency)Features=–2026.0633.02
Random ChoiceCategory=Reference2026.0533
Random (frequency)Features=–2026.0632.66
GPT-4oFeatures=Passive2026.0632.65
Random (chance)Features=–2026.0632.35
Random (chance)Features=–2026.0632.29
Spatial-MLLMFeatures=Passive, 2D+3D2026.0632.06
Video-R1Training Protocol=RL2025.1231.9
Qwen2.5-VL-7B-InstructFeatures=Passive2026.0629.5
Qwen3-VL-8B-InstructCategory=Open-source General Models2026.0529.4
Qwen2.5-VL-7B-InstructFeatures=Passive2026.0629.26
Spatial-MLLMFeatures=Passive, 2D+3D2026.0620.92