List your company
ResearchFormerly WizWand

Spatial Reasoning on SpatialReasonerEval

92.6AccuracySpatialThinker-30B
59.94468.42276.985.378Nov 10, 2025
Updated 3mo ago

Evaluation results

MethodLinks
SpatialThinker-30BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1192.6
GPT-5-0807Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1190.5
Qwen3-VL-30BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1188.2
GPT-4o-0513Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1185.8
Claude-4-Sonnet-0514Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1185.7
Claude-3.5-Sonnet-0620Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1184.1
SpatialThinker-7BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1182.7
Qwen2.5-VL-7B + Vanilla GRPOEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1179.6
SpatialThinker-3BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1176.5
SpatialReasonerEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1176.4
Visionary-R1Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1172.9
Qwen2.5-VL-7B + SFTEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1170.8
Qwen2.5-VL-7BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1170.6
SATORI-R1Evaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1170.5
SpaceThinkerEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1169.6
Qwen2.5-VL-3B + Vanilla GRPOEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1169.3
SpaceOmEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1168.9
Qwen2.5-VL-3BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1168
Qwen2.5-VL-3B + SFTEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB, Training Dataset=STVQA-7K2025.1167.5
VLAA-Thinker-7BEvaluation Setting=Zero-shot, Decoding=Greedy, Input Modality=RGB2025.1161.2