Share your thoughts, 1 month free Claude Pro on usSee more

Geometry Problem Solving on OlympiadBench

75.5Accuracy

GPT-5

Updated 1mo ago

Evaluation Results

Method	Links
GPT-5 2026.02		75.5
Gemini-2.5-Pro 2026.02		75.22
Qwen3-VL-7B-Instruct (GeoCode trained) 2026.02		57.34
Qwen3-VL 2026.02		54.46
Qwen3-VL-7B-Instruct (Baseline) 2026.02		45.82
GeometryZero-VL-7B 2025.06		37.19
Qwen2.5-VL-7B-Instruct + ToRL 2025.06		34.59
Qwen2.5-VL-7B-Instruct + SFT 2025.06		32.36
Qwen2.5-VL-7B-Instruct + GRPO 2025.06		31.82
Qwen2.5-VL-7B-Instruct 2025.06		30.83
Qwen2.5-VL-7B-Instruct (GeoCode trained) 2026.02		6.63
Qwen2.5-VL-7B-Instruct (Baseline) 2026.02		5.76