# Mathematical Reasoning on AIME 24 (Accuracy (%)) Source: [Mathematical Reasoning on AIME 24 (Accuracy (%))]() Collection index: [llms.txt]() Data updated: 2026-02-26T05:30:50.033Z Evaluation of model performance on the AIME 2024 dataset for mathematical reasoning, specifically reporting the Accuracy metric as a percentage. 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) | AIME 24 | | Metric | 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 - [Accuracy](): Higher is better; unit: PERCENT. ## Recorded results | Method | Accuracy | Result date | Report paper | | --- | --- | --- | --- | | ELPO-4B | 74.3 | Not recorded | 698c0d54c97a0b3e3b5e136c | | DemyAgent-4B | 72.6 | Not recorded | 698c0d54c97a0b3e3b5e136c | | Qwen3-4B | 63.3 | Not recorded | 698c0d54c97a0b3e3b5e136c | | ELPO-7B | 48.1 | Not recorded | 698c0d54c97a0b3e3b5e136c | | DemyAgent | 43.2 | Not recorded | 698c0d54c97a0b3e3b5e136c | | CIR | 42.3 | Not recorded | 698c0d54c97a0b3e3b5e136c | | ToRL | 40.2 | Not recorded | 698c0d54c97a0b3e3b5e136c | | DeepSeek-V3 | 39.2 | Not recorded | 698c0d54c97a0b3e3b5e136c | | AEPO | 33.3 | Not recorded | 698c0d54c97a0b3e3b5e136c | | ARPO | 30 | Not recorded | 698c0d54c97a0b3e3b5e136c | | GIGPO | 30 | Not recorded | 698c0d54c97a0b3e3b5e136c | | Reinforce++ | 26.7 | Not recorded | 698c0d54c97a0b3e3b5e136c | | GRPO | 23.3 | Not recorded | 698c0d54c97a0b3e3b5e136c | | DAPO | 20 | Not recorded | 698c0d54c97a0b3e3b5e136c | | Qwen3-4B | 17 | Not recorded | 698c0d54c97a0b3e3b5e136c | | Qwen2.5-7B | 16.7 | Not recorded | 698c0d54c97a0b3e3b5e136c | | Qwen2.5-7B | 6.7 | Not recorded | 698c0d54c97a0b3e3b5e136c | ### ELPO-4B Accuracy: 74.3. Result ID: 69963d565698665ec4806f11. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Backbone Architecture Family":"Qwen3-4B"}. This record does not assert independent reproduction. ### DemyAgent-4B Accuracy: 72.6. Result ID: 69963d565698665ec4806f10. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Backbone Architecture Family":"Qwen3-4B"}. This record does not assert independent reproduction. ### Qwen3-4B Accuracy: 63.3. Result ID: 69963d565698665ec4806f02. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Reasoning Strategy":"Self-Contained Reasoning","Backbone Architecture Family":"Qwen3-4B"}. This record does not assert independent reproduction. ### ELPO-7B Accuracy: 48.1. Result ID: 69963d565698665ec4806f0f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Optimization/RL Method Category":"Agentic RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### DemyAgent Accuracy: 43.2. Result ID: 69963d565698665ec4806f0e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Optimization/RL Method Category":"Agentic RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### CIR Accuracy: 42.3. Result ID: 69963d565698665ec4806f0c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Math","Optimization/RL Method Category":"Agentic RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### ToRL Accuracy: 40.2. Result ID: 69963d565698665ec4806f09. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Math-Inst","Optimization/RL Method Category":"Agentic RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### DeepSeek-V3 Accuracy: 39.2. Result ID: 69963d565698665ec4806f03. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Base","Reasoning Strategy":"Self-Contained Reasoning","Backbone Architecture Family":"DeepSeek-V3"}. This record does not assert independent reproduction. ### AEPO Accuracy: 33.3. Result ID: 69963d565698665ec4806f0b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Optimization/RL Method Category":"Agentic RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### ARPO Accuracy: 30. Result ID: 69963d565698665ec4806f0a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Optimization/RL Method Category":"Agentic RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### GIGPO Accuracy: 30. Result ID: 69963d565698665ec4806f0d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Optimization/RL Method Category":"Agentic RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### Reinforce++ Accuracy: 26.7. Result ID: 69963d565698665ec4806f07. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Instruct","Optimization/RL Method Category":"Classical RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### GRPO Accuracy: 23.3. Result ID: 69963d565698665ec4806f06. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Optimization/RL Method Category":"Classical RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### DAPO Accuracy: 20. Result ID: 69963d565698665ec4806f08. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Instruct","Optimization/RL Method Category":"Clipping-optimized RL","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### Qwen3-4B Accuracy: 17. Result ID: 69963d565698665ec4806f05. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Reasoning Strategy":"TIR Reasoning","Backbone Architecture Family":"Qwen3-4B"}. This record does not assert independent reproduction. ### Qwen2.5-7B Accuracy: 16.7. Result ID: 69963d565698665ec4806f01. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Reasoning Strategy":"Self-Contained Reasoning","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. ### Qwen2.5-7B Accuracy: 6.7. Result ID: 69963d565698665ec4806f04. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Starting Checkpoint":"Inst","Reasoning Strategy":"TIR Reasoning","Backbone Architecture Family":"Qwen2.5-7B"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.