# Geometry Problem Solving on OlympiadBench (BoN@3 Accuracy) Source: [Geometry Problem Solving on OlympiadBench (BoN@3 Accuracy)]() Collection index: [llms.txt]() Data updated: 2026-04-21T12:56:27.515Z Evaluation of model performance on geometry problem solving tasks using the OlympiadBench dataset, specifically measured by Best-of-N (N=3) accuracy. A research result applies to its stated dataset, split, metric and evaluation configuration. Compare like-for-like settings, preserve the metric's unit and direction, and distinguish paper-reported results from independent reproduction. The best result recorded here is not a claim of exhaustive, current global coverage. | Scope | Recorded value | | --- | --- | | Task | Geometry problem solving | | Dataset (including any stated split) | OlympiadBench | | Metric | BoN@3 Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 17 of 17 | | Coverage | All results currently returned for this benchmark | Evaluation settings may differ between results. The original papers and recorded configurations are needed to establish comparability. A shared metric name alone does not establish an identical protocol. ## Metrics - [BoN@3 Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | BoN@3 Accuracy | Result date | Report paper | | --- | --- | --- | --- | | GeometryZero-7B | 45.69 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-7B-Instruct + ToRL | 44.08 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-7B-Instruct + SFT | 41.93 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-7B-Instruct + GRPO | 40.32 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-7B-Instruct | 39.24 | Not recorded | 68fa79d01e209131cea7f82e | | GeometryZero-3B | 35.48 | Not recorded | 68fa79d01e209131cea7f82e | | GNS-Llava-1.5-7B | 33.54 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-3B-Instruct + ToRL | 32.87 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-3B-Instruct | 32.25 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-3B-Instruct + GRPO | 31.72 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-3B-Instruct + SFT | 30.64 | Not recorded | 68fa79d01e209131cea7f82e | | G-Llava-7B | 27.82 | Not recorded | 68fa79d01e209131cea7f82e | | GeometryZero-1.5B | 16.47 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-1.5B-Instruct + SFT | 14.51 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-1.5B-Instruct + GRPO | 14.51 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-1.5B-Instruct | 13.44 | Not recorded | 68fa79d01e209131cea7f82e | | Qwen2.5-1.5B-Instruct + ToRL | 11.29 | Not recorded | 68fa79d01e209131cea7f82e | ### GeometryZero-7B BoN@3 Accuracy: 45.69. Result ID: 69e772de47f34ce924cf321c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"7B"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct + ToRL BoN@3 Accuracy: 44.08. Result ID: 69e772de47f34ce924cf321b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"7B"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct + SFT BoN@3 Accuracy: 41.93. Result ID: 69e772de47f34ce924cf3219. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"7B"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct + GRPO BoN@3 Accuracy: 40.32. Result ID: 69e772de47f34ce924cf321a. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Size":"7B"}. This record does not assert independent reproduction. ### Qwen2.5-7B-Instruct BoN@3 Accuracy: 39.24. Result ID: 69e772de47f34ce924cf3218. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"7B"}. This record does not assert independent reproduction. ### GeometryZero-3B BoN@3 Accuracy: 35.48. Result ID: 69e772de47f34ce924cf3215. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction. ### GNS-Llava-1.5-7B BoN@3 Accuracy: 33.54. Result ID: 69e772de47f34ce924cf3217. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"7B"}. This record does not assert independent reproduction. ### Qwen2.5-3B-Instruct + ToRL BoN@3 Accuracy: 32.87. Result ID: 69e772de47f34ce924cf3214. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction. ### Qwen2.5-3B-Instruct BoN@3 Accuracy: 32.25. Result ID: 69e772de47f34ce924cf3211. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction. ### Qwen2.5-3B-Instruct + GRPO BoN@3 Accuracy: 31.72. Result ID: 69e772de47f34ce924cf3213. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction. ### Qwen2.5-3B-Instruct + SFT BoN@3 Accuracy: 30.64. Result ID: 69e772de47f34ce924cf3212. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"3B"}. This record does not assert independent reproduction. ### G-Llava-7B BoN@3 Accuracy: 27.82. Result ID: 69e772de47f34ce924cf3216. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"7B"}. This record does not assert independent reproduction. ### GeometryZero-1.5B BoN@3 Accuracy: 16.47. Result ID: 69e772de47f34ce924cf3210. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1.5B"}. This record does not assert independent reproduction. ### Qwen2.5-1.5B-Instruct + SFT BoN@3 Accuracy: 14.51. Result ID: 69e772de47f34ce924cf320d. [Reporting paper]() Recorded evaluation configuration: {"Model Size":"1.5B"}. This record does not assert independent reproduction. ### Qwen2.5-1.5B-Instruct + GRPO BoN@3 Accuracy: 14.51. Result ID: 69e772de47f34ce924cf320e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Size":"1.5B"}. This record does not assert independent reproduction. ### Qwen2.5-1.5B-Instruct BoN@3 Accuracy: 13.44. Result ID: 69e772de47f34ce924cf320c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1.5B"}. This record does not assert independent reproduction. ### Qwen2.5-1.5B-Instruct + ToRL BoN@3 Accuracy: 11.29. Result ID: 69e772de47f34ce924cf320f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"1.5B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.