# Mathematical Reasoning on MATH (SC/ORM/PRM Metrics) Source: [Mathematical Reasoning on MATH (SC/ORM/PRM Metrics)]() Collection index: [llms.txt]() Data updated: 2026-03-04T15:12:38.778Z Evaluation of model performance on the MATH dataset for complex mathematical reasoning, utilizing specific metrics such as SC (Standard Correctness), ORM+SC, PRM+SC, and ΔRM. 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 | | Metric | Standard Correctness (SC) | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 9 of 9 | | 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 - [Standard Correctness (SC)](): Higher is better; unit: PERCENT. - [ORM + SC](): Higher is better; unit: PERCENT. - [PRM + SC](): Higher is better; unit: PERCENT. - [Delta RM (ΔRM)](): Higher is better; unit: SCALAR. ## Recorded results | Method | Standard Correctness (SC) | Result date | Report paper | | --- | --- | --- | --- | | SPARE-Qwen2.5-3B | 66.6 | Not recorded | 68fa7ad71e209131cea8574d | | SPARE-Qwen2.5-3B | 64.6 | Not recorded | 68fa7ad71e209131cea8574d | | Math-Shepherd | 45.4 | Not recorded | 68fa7ad71e209131cea8574d | | Math-Shepherd | 42.3 | Not recorded | 68fa7ad71e209131cea8574d | | Math-Shepherd | 35.1 | Not recorded | 68fa7ad71e209131cea8574d | | Rest-MCTS\* | 35.1 | Not recorded | 68fa7ad71e209131cea8574d | | SPARE-Qwen2.5-3B | 31.4 | Not recorded | 68fa7ad71e209131cea8574d | | SPARE-Llama3-8B | 30.5 | Not recorded | 68fa7ad71e209131cea8574d | | SPARE-Llama3-8B | 23.4 | Not recorded | 68fa7ad71e209131cea8574d | ### SPARE-Qwen2.5-3B Standard Correctness (SC): 66.6. Result ID: 69a84bdf6dbb7075b94f565b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N":"20","Generator":"Qwen2.5-3B Gen.","Generation length":"2048","Note":"using pre-trained model"}. This record does not assert independent reproduction. ### SPARE-Qwen2.5-3B Standard Correctness (SC): 64.6. Result ID: 69a84bdf6dbb7075b94f565c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N":"20","Generator":"Qwen2.5-32B Gen."}. This record does not assert independent reproduction. ### Math-Shepherd Standard Correctness (SC): 45.4. Result ID: 69a84bdf6dbb7075b94f565f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"DeepSeek-67B","N":"256"}. This record does not assert independent reproduction. ### Math-Shepherd Standard Correctness (SC): 42.3. Result ID: 69a84bdf6dbb7075b94f565e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Mistral-7B","Training paradigm":"Process RL Gen.","N":"256"}. This record does not assert independent reproduction. ### Math-Shepherd Standard Correctness (SC): 35.1. Result ID: 69a84bdf6dbb7075b94f565d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Mistral-7B","Training paradigm":"SFT Gen.","N":"256"}. This record does not assert independent reproduction. ### Rest-MCTS\* Standard Correctness (SC): 35.1. Result ID: 69a84bdf6dbb7075b94f5660. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Mistral-7B","N":"256"}. This record does not assert independent reproduction. ### SPARE-Qwen2.5-3B Standard Correctness (SC): 31.4. Result ID: 69a84bdf6dbb7075b94f565a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N":"20","Generator":"Qwen2.5-3B Gen."}. This record does not assert independent reproduction. ### SPARE-Llama3-8B Standard Correctness (SC): 30.5. Result ID: 69a84bdf6dbb7075b94f5659. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N":"256"}. This record does not assert independent reproduction. ### SPARE-Llama3-8B Standard Correctness (SC): 23.4. Result ID: 69a84bdf6dbb7075b94f5658. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"N":"20"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.