List your company

Spatial Reasoning on VSR (ood)

84.8AccuracyThinking with Blueprint
76.16878.40980.6582.891Jan 5, 2026

Evaluation results

MethodLinks
Thinking with BlueprintModel Size=7B, Reasoning Mode=Thinking with Blueprint2026.0184.8
gpt-5-ThinkingModel Size=-, Reasoning Mode=Naive CoT2026.0184.2
RobixModel Size=32B, Reasoning Mode=Naive CoT2026.0183.7
Qwen2.5-VLModel Size=7B, Reasoning Mode=No2026.0183.6
RobixModel Size=7B, Reasoning Mode=Naive CoT2026.0183.3
VigoRLModel Size=7B, Reasoning Mode=Thinking with Coordinates2026.0182.7
gpt-4oModel Size=-, Reasoning Mode=Naive CoT2026.0182.2
Qwen2.5-VLModel Size=7B, Reasoning Mode=Naive CoT2026.0182.2
gpt-4oModel Size=-, Reasoning Mode=No2026.0178.7
ViLASRModel Size=7B, Reasoning Mode=Thinking with Images2026.0176.5