# Mathematical Reasoning on MATH-500 (Accuracy, Delta, Token Length, Ratio) Source: [Mathematical Reasoning on MATH-500 (Accuracy, Delta, Token Length, Ratio)]() Collection index: [llms.txt]() Data updated: 2026-05-11T10:03:04.777Z Evaluation of model performance on the MATH-500 dataset for mathematical reasoning, reporting standard accuracy, a delta score, average token length, and an average ratio metric. 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) | MATH-500 | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 19 of 19 | | 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 - [Accuracy](): Higher is better; unit: PERCENT. - [Delta Score](): Higher is better; unit: SCALAR. - [Average Token Length](): Higher is better; unit: SCALAR. - [Average Ratio](): Higher is better; unit: SCALAR. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Stability filtering + Content projection | 81.6 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | Stability filtering + Content projection | 81.2 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | IRIS | 80.5 | Not recorded | 6a0162b8dbc836b87d8232fd | | IRIS | 79.5 | Not recorded | 6a0162b8dbc836b87d8232fd | | IRIS | 79.5 | Not recorded | 6a0162b8dbc836b87d8232fd | | IRIS | 78.5 | Not recorded | 6a0162b8dbc836b87d8232fd | | Stability filtering + Content projection | 78.4 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | SEAL | 76.6 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | SEAL | 75.2 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | SEAL | 73.4 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | Baseline | 72.6 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | Prompt text | 67.2 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | Baseline | 66.2 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | Prompt vector | 61 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | Baseline | 60.8 | Not recorded | 69cf49bd0636e5bf4afcdbd1 | | TOWN | 47 | Not recorded | 6a0162b8dbc836b87d8232fd | | TOWN | 45 | Not recorded | 6a0162b8dbc836b87d8232fd | | TOWN | 44 | Not recorded | 6a0162b8dbc836b87d8232fd | | TOWN | 44 | Not recorded | 6a0162b8dbc836b87d8232fd | ### Stability filtering + Content projection Accuracy: 81.6. Result ID: 69cf83ba47f34ce924b99f2b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Nemotron-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Cross-model transfer","Steering vector source":"DeepSeek-R1-Distill-Qwen-1.5B"}. This record does not assert independent reproduction. ### Stability filtering + Content projection Accuracy: 81.2. Result ID: 69cf83ba47f34ce924b99f2e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"DeepScaleR-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Cross-model transfer","Steering vector source":"DeepSeek-R1-Distill-Qwen-1.5B"}. This record does not assert independent reproduction. ### IRIS Accuracy: 80.5. Result ID: 6a01a95847f34ce924e40f65. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seed":"456","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. ### IRIS Accuracy: 79.5. Result ID: 6a01a95847f34ce924e40f61. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seed":"42","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. ### IRIS Accuracy: 79.5. Result ID: 6a01a95847f34ce924e40f67. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seed":"Mean","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. ### IRIS Accuracy: 78.5. Result ID: 6a01a95847f34ce924e40f63. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Seed":"123","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. ### Stability filtering + Content projection Accuracy: 78.4. Result ID: 69cf83ba47f34ce924b99f28. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Source model","Steering vector source":"DeepSeek-R1-Distill-Qwen-1.5B"}. This record does not assert independent reproduction. ### SEAL Accuracy: 76.6. Result ID: 69cf83ba47f34ce924b99f2a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Nemotron-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Cross-model transfer","Steering vector source":"DeepSeek-R1-Distill-Qwen-1.5B"}. This record does not assert independent reproduction. ### SEAL Accuracy: 75.2. Result ID: 69cf83ba47f34ce924b99f2d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"DeepScaleR-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Cross-model transfer","Steering vector source":"DeepSeek-R1-Distill-Qwen-1.5B"}. This record does not assert independent reproduction. ### SEAL Accuracy: 73.4. Result ID: 69cf83ba47f34ce924b99f27. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model":"Qwen-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Source model","Steering vector source":"DeepSeek-R1-Distill-Qwen-1.5B"}. This record does not assert independent reproduction. ### Baseline Accuracy: 72.6. Result ID: 69cf83ba47f34ce924b99f2c. [Reporting paper]() Recorded evaluation configuration: {"Model":"DeepScaleR-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Cross-model transfer"}. This record does not assert independent reproduction. ### Prompt text Accuracy: 67.2. Result ID: 69cf83ba47f34ce924b99f25. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Source model"}. This record does not assert independent reproduction. ### Baseline Accuracy: 66.2. Result ID: 69cf83ba47f34ce924b99f29. [Reporting paper]() Recorded evaluation configuration: {"Model":"Nemotron-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Cross-model transfer"}. This record does not assert independent reproduction. ### Prompt vector Accuracy: 61. Result ID: 69cf83ba47f34ce924b99f26. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Source model"}. This record does not assert independent reproduction. ### Baseline Accuracy: 60.8. Result ID: 69cf83ba47f34ce924b99f24. [Reporting paper]() Recorded evaluation configuration: {"Model":"Qwen-1.5B","max\_tokens":"4096","layer":"20","Transfer setting":"Source model"}. This record does not assert independent reproduction. ### TOWN Accuracy: 47. Result ID: 6a01a95847f34ce924e40f66. [Reporting paper]() Recorded evaluation configuration: {"Seed":"456","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. ### TOWN Accuracy: 45. Result ID: 6a01a95847f34ce924e40f68. [Reporting paper]() Recorded evaluation configuration: {"Seed":"Mean","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. ### TOWN Accuracy: 44. Result ID: 6a01a95847f34ce924e40f62. [Reporting paper]() Recorded evaluation configuration: {"Seed":"42","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. ### TOWN Accuracy: 44. Result ID: 6a01a95847f34ce924e40f64. [Reporting paper]() Recorded evaluation configuration: {"Seed":"123","B\_think":"4096","B\_answer":"512","n":"200","Hardware":"A100"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.