# Multi-modal Reasoning on MathVista (Accuracy) Source: [Multi-modal Reasoning on MathVista (Accuracy)]() Collection index: [llms.txt]() Data updated: 2026-05-11T07:10:15.757Z Evaluation of model performance on the MathVista dataset for Multi-modal Reasoning, specifically reporting the Accuracy 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 | Multi-modal Reasoning | | Dataset (including any stated split) | MathVista | | Metric | Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 16 of 16 | | 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 | | --- | --- | --- | --- | | AutoNPO | 79.2 | Not recorded | 69e9a7bd55a3c32af89f7266 | | RLEP | 78.5 | Not recorded | 69e9a7bd55a3c32af89f7266 | | ExGRPO | 77.3 | Not recorded | 69e9a7bd55a3c32af89f7266 | | NPO | 76.6 | Not recorded | 69e9a7bd55a3c32af89f7266 | | NPO | 76.3 | Not recorded | 69e9a7bd55a3c32af89f7266 | | GRPO | 76.2 | Not recorded | 69e9a7bd55a3c32af89f7266 | | Qwen3-VL-8B-Instruct | 73.8 | Not recorded | 69e9a7bd55a3c32af89f7266 | | LUFFY | 73.8 | Not recorded | 69e9a7bd55a3c32af89f7266 | | Self-Instruct + Solver Feedback | 70.3 | Not recorded | 6916b7990052cac1dc65e123 | | CoT Cold-Start + Solver Feedback | 70.1 | Not recorded | 6916b7990052cac1dc65e123 | | Self-Instruct | 69.8 | Not recorded | 6916b7990052cac1dc65e123 | | Self-Instruct + CoT-Self-Instruct | 69.3 | Not recorded | 6916b7990052cac1dc65e123 | | Self-Instruct + R-Zero | 69.3 | Not recorded | 6916b7990052cac1dc65e123 | | CoT Cold-Start | 69.1 | Not recorded | 6916b7990052cac1dc65e123 | | Seed Set | 69 | Not recorded | 6916b7990052cac1dc65e123 | | Base Model | 68.4 | Not recorded | 6916b7990052cac1dc65e123 | ### AutoNPO Accuracy: 79.2. Result ID: 69e9fe6b47f34ce924d22233. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### RLEP Accuracy: 78.5. Result ID: 69e9fe6b47f34ce924d22230. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"type":"far future"}. This record does not assert independent reproduction. ### ExGRPO Accuracy: 77.3. Result ID: 69e9fe6b47f34ce924d2222f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"type":"historical replay"}. This record does not assert independent reproduction. ### NPO Accuracy: 76.6. Result ID: 69e9fe6b47f34ce924d22231. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"stage":"early-stage only"}. This record does not assert independent reproduction. ### NPO Accuracy: 76.3. Result ID: 69e9fe6b47f34ce924d22232. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"stage":"early + late-stage"}. This record does not assert independent reproduction. ### GRPO Accuracy: 76.2. Result ID: 69e9fe6b47f34ce924d2222e. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"type":"pure on-policy"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Accuracy: 73.8. Result ID: 69e9fe6b47f34ce924d2222c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: Not recorded. This record does not assert independent reproduction. ### LUFFY Accuracy: 73.8. Result ID: 69e9fe6b47f34ce924d2222d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"type":"external teacher"}. This record does not assert independent reproduction. ### Self-Instruct + Solver Feedback Accuracy: 70.3. Result ID: 6a0180c547f34ce924e2dc5e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. ### CoT Cold-Start + Solver Feedback Accuracy: 70.1. Result ID: 6a0180c547f34ce924e2dc60. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. ### Self-Instruct Accuracy: 69.8. Result ID: 6a0180c547f34ce924e2dc5b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. ### Self-Instruct + CoT-Self-Instruct Accuracy: 69.3. Result ID: 6a0180c547f34ce924e2dc5c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. ### Self-Instruct + R-Zero Accuracy: 69.3. Result ID: 6a0180c547f34ce924e2dc5d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. ### CoT Cold-Start Accuracy: 69.1. Result ID: 6a0180c547f34ce924e2dc5f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. ### Seed Set Accuracy: 69. Result ID: 6a0180c547f34ce924e2dc5a. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. ### Base Model Accuracy: 68.4. Result ID: 6a0180c547f34ce924e2dc59. [Reporting paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B-Instruct"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.