List your company

Geometry Problem Solving on UniGeo calculation (test)

96.9Top-1 AccuracyDraw2Think CT
40.11654.85869.684.342May 20, 2026

Evaluation results

MethodLinks
Draw2Think CTBackbone=gemini-3-flash-preview, Execution Environment=executable environment2026.0596.9
Draw2Think BLBackbone=gemini-3-flash-preview, Execution Environment=direct VLM2026.0593.6
Gemini-2.5-ProProtocol=Zero-shot, Reported by=TrustGeoGen [9]2026.0579.6
OpenAI-o3Protocol=Zero-shot, Reported by=TrustGeoGen [9]2026.0579.3
GeoGen-SFTBackbone=qwen2.5-vl-7b, Protocol=Fine-tuned, Mechanism=SFT2026.0578
GeoSketch-RLBackbone=qwen2.5-vl-7b, Protocol=Fine-tuned, Mechanism=SFT+RL2026.0572.5
NeSyGeoBackbone=qwen2.5-vl-7b, Protocol=Fine-tuned, Mechanism=SFT2026.0571.8
GeoDRLProtocol=Neural-symbolic, Input Format=Parsed diagram annotations2026.0566.7
Qwen2.5-VL-7BProtocol=Zero-shot, Reported by=GeoGen [35]2026.0564.2
SocraticGeoBackbone=qwen2.5-vl-7b, Protocol=Fine-tuned, Mechanism=GRPO2026.0549.2
GPT-4oProtocol=Zero-shot, Reported by=TrustGeoGen [9]2026.0542.3