List your company

Geometric Problem Solving on GeoSketch 1.0 (test)

86.14Total AccuracyHuman
30.416844.883459.3573.8166Sep 26, 2025

Evaluation results

MethodLinks
HumanEvaluation Framework=Pure Run (Baseline)2025.0986.1484.0887.0490.12
GeoSketch-Qwen2.5VL-7B-RLEvaluation Framework=GeoSketch Framework2025.0959.2357.2160.1962.96
Qwen2.5VL-72BEvaluation Framework=GeoSketch Framework2025.0958.7157.7156.4864.2
GeoSketch-Qwen2.5VL-7B-SFTEvaluation Framework=GeoSketch Framework2025.0952.5649.7560.1949.38
GPT-4oEvaluation Framework=GeoSketch Framework2025.0949.7444.2859.2650.62
Qwen2.5VL-7BEvaluation Framework=GeoSketch Framework2025.0948.4645.7757.443.21
Qwen2.5VL-72BEvaluation Framework=Pure Run (Baseline)2025.0943.0838.8149.0745.68
GeoSketch-Qwen2.5VL-7B-RLEvaluation Framework=Pure Run (Baseline)2025.0935.6428.3648.1537.04
GeoSketch-Qwen2.5VL-7B-SFTEvaluation Framework=Pure Run (Baseline)2025.0933.8525.3750.9332.1
GPT-4oEvaluation Framework=Pure Run (Baseline)2025.0933.0824.885030.86
Qwen2.5VL-7BEvaluation Framework=Pure Run (Baseline)2025.0932.5623.885030.86