# Mathematical Reasoning on Olympiad Bench (Peak Accuracy) Source: [Mathematical Reasoning on Olympiad Bench (Peak Accuracy)]() Collection index: [llms.txt]() Data updated: 2026-05-27T08:40:39.435Z Evaluation of model performance on mathematical reasoning problems using the Olympiad Bench dataset, specifically measured using Peak 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 | Mathematical Reasoning | | Dataset (including any stated split) | Olympiad Bench | | Metric | Peak Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 12 of 12 | | 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 - [Peak Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Peak Accuracy | Result date | Report paper | | --- | --- | --- | --- | | 14B-PC | 0.57 | Not recorded | 6a167a79dbd7150506f7fed0 | | 14B-GRPO | 0.567 | Not recorded | 6a167a79dbd7150506f7fed0 | | 14B-DAPO | 0.527 | Not recorded | 6a167a79dbd7150506f7fed0 | | 8B-DAPO | 0.525 | Not recorded | 6a167a79dbd7150506f7fed0 | | 8B-PC | 0.507 | Not recorded | 6a167a79dbd7150506f7fed0 | | 4B-PC | 0.497 | Not recorded | 6a167a79dbd7150506f7fed0 | | 4B-DAPO | 0.476 | Not recorded | 6a167a79dbd7150506f7fed0 | | 4B-GRPO | 0.47 | Not recorded | 6a167a79dbd7150506f7fed0 | | 8B-GRPO | 0.457 | Not recorded | 6a167a79dbd7150506f7fed0 | | 1.5B-DAPO | 0.43 | Not recorded | 6a167a79dbd7150506f7fed0 | | 1.5B-PC | 0.43 | Not recorded | 6a167a79dbd7150506f7fed0 | | 1.5B-GRPO | 0.418 | Not recorded | 6a167a79dbd7150506f7fed0 | ### 14B-PC Peak Accuracy: 0.57. Result ID: 6a16ace4fbc373592c0fd8f6. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"14B","RL Algorithm":"PC","Rollouts Trained":"2,048","Rollouts Sampled":"4,096"}. This record does not assert independent reproduction. ### 14B-GRPO Peak Accuracy: 0.567. Result ID: 6a16ace4fbc373592c0fd8f4. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Scale":"14B","RL Algorithm":"GRPO","Rollouts Trained":"2,048","Rollouts Sampled":"2,048"}. This record does not assert independent reproduction. ### 14B-DAPO Peak Accuracy: 0.527. Result ID: 6a16ace4fbc373592c0fd8f5. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"14B","RL Algorithm":"DAPO","Rollouts Trained":"2,048","Rollouts Sampled":"6,144"}. This record does not assert independent reproduction. ### 8B-DAPO Peak Accuracy: 0.525. Result ID: 6a16ace4fbc373592c0fd8f2. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"8B","RL Algorithm":"DAPO","Rollouts Trained":"2,048","Rollouts Sampled":"6,144"}. This record does not assert independent reproduction. ### 8B-PC Peak Accuracy: 0.507. Result ID: 6a16ace4fbc373592c0fd8f3. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"8B","RL Algorithm":"PC","Rollouts Trained":"2,048","Rollouts Sampled":"4,096"}. This record does not assert independent reproduction. ### 4B-PC Peak Accuracy: 0.497. Result ID: 6a16ace4fbc373592c0fd8f0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"4B","RL Algorithm":"PC","Rollouts Trained":"2,048","Rollouts Sampled":"4,096"}. This record does not assert independent reproduction. ### 4B-DAPO Peak Accuracy: 0.476. Result ID: 6a16ace4fbc373592c0fd8ef. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"4B","RL Algorithm":"DAPO","Rollouts Trained":"2,048","Rollouts Sampled":"6,144"}. This record does not assert independent reproduction. ### 4B-GRPO Peak Accuracy: 0.47. Result ID: 6a16ace4fbc373592c0fd8ee. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Scale":"4B","RL Algorithm":"GRPO","Rollouts Trained":"2,048","Rollouts Sampled":"2,048"}. This record does not assert independent reproduction. ### 8B-GRPO Peak Accuracy: 0.457. Result ID: 6a16ace4fbc373592c0fd8f1. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Scale":"8B","RL Algorithm":"GRPO","Rollouts Trained":"2,048","Rollouts Sampled":"2,048"}. This record does not assert independent reproduction. ### 1.5B-DAPO Peak Accuracy: 0.43. Result ID: 6a16ace4fbc373592c0fd8ec. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"1.5B","RL Algorithm":"DAPO","Rollouts Trained":"2,048","Rollouts Sampled":"6,144"}. This record does not assert independent reproduction. ### 1.5B-PC Peak Accuracy: 0.43. Result ID: 6a16ace4fbc373592c0fd8ed. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Scale":"1.5B","RL Algorithm":"PC","Rollouts Trained":"2,048","Rollouts Sampled":"4,096"}. This record does not assert independent reproduction. ### 1.5B-GRPO Peak Accuracy: 0.418. Result ID: 6a16ace4fbc373592c0fd8eb. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Scale":"1.5B","RL Algorithm":"GRPO","Rollouts Trained":"2,048","Rollouts Sampled":"2,048"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.