# Reasoning and Math on VLMEvalKit (test) Source: [Reasoning and Math on VLMEvalKit (test)]() Collection index: [llms.txt]() Data updated: 2026-03-05T07:07:22.584Z Evaluation of model performance on the test split of the VLMEvalKit suite, aggregating results across various mathematical reasoning and logic sub-evaluations like MathVista, MathVision, and WeMath. 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 | Reasoning and Math | | Dataset (including any stated split) | VLMEvalKit (test) | | Metric | MathVista Accuracy | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 13 of 13 | | 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 - [MathVista Accuracy](): Higher is better; unit: PERCENT. - [MathVision Accuracy](): Higher is better; unit: PERCENT. - [WeMath Accuracy](): Higher is better; unit: PERCENT. - [MathVerse Accuracy](): Higher is better; unit: PERCENT. - [LogicVista Accuracy](): Higher is better; unit: PERCENT. - [DynaMath Accuracy](): Higher is better; unit: PERCENT. - [Average Accuracy (VLMEvalKit)](): Higher is better; unit: PERCENT. ## Recorded results | Method | MathVista Accuracy | Result date | Report paper | | --- | --- | --- | --- | | Gemini2.0-Flash | 73.4 | Not recorded | 68fa7bb11e209131cea89eff | | VisionZero-Qwen-7B (Real-World) | 73.1 | Not recorded | 68fa7bb11e209131cea89eff | | MM-Eureka-Qwen-7B | 73 | Not recorded | 68fa7bb11e209131cea89eff | | VisionZero-Qwen-7B (Chart) | 72.4 | Not recorded | 68fa7bb11e209131cea89eff | | VisionZero-Qwen-7B (CLEVR) | 72.2 | Not recorded | 68fa7bb11e209131cea89eff | | ViGaL-Snake+Rotation | 71.9 | Not recorded | 68fa7bb11e209131cea89eff | | ViGaL-Rotation | 71.2 | Not recorded | 68fa7bb11e209131cea89eff | | ViGaL-Snake | 70.7 | Not recorded | 68fa7bb11e209131cea89eff | | OpenVLThinker-7B | 70.2 | Not recorded | 68fa7bb11e209131cea89eff | | Qwen2.5-VL-7B | 68.2 | Not recorded | 68fa7bb11e209131cea89eff | | VLAA-Thinker-7B | 68 | Not recorded | 68fa7bb11e209131cea89eff | | R1-OneVision-7B | 64.1 | Not recorded | 68fa7bb11e209131cea89eff | | GPT4o | 61.4 | Not recorded | 68fa7bb11e209131cea89eff | ### Gemini2.0-Flash MathVista Accuracy: 73.4. Result ID: 69a92b8a6dbb7075b9520d30. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Category":"Proprietary Model"}. This record does not assert independent reproduction. ### VisionZero-Qwen-7B (Real-World) MathVista Accuracy: 73.1. Result ID: 69a92b8a6dbb7075b9520d3b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","Training Domain":"Real-World"}. This record does not assert independent reproduction. ### MM-Eureka-Qwen-7B MathVista Accuracy: 73. Result ID: 69a92b8a6dbb7075b9520d33. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### VisionZero-Qwen-7B (Chart) MathVista Accuracy: 72.4. Result ID: 69a92b8a6dbb7075b9520d3a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","Training Domain":"Chart"}. This record does not assert independent reproduction. ### VisionZero-Qwen-7B (CLEVR) MathVista Accuracy: 72.2. Result ID: 69a92b8a6dbb7075b9520d39. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B","Training Domain":"CLEVR"}. This record does not assert independent reproduction. ### ViGaL-Snake+Rotation MathVista Accuracy: 71.9. Result ID: 69a92b8a6dbb7075b9520d38. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### ViGaL-Rotation MathVista Accuracy: 71.2. Result ID: 69a92b8a6dbb7075b9520d37. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### ViGaL-Snake MathVista Accuracy: 70.7. Result ID: 69a92b8a6dbb7075b9520d36. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### OpenVLThinker-7B MathVista Accuracy: 70.2. Result ID: 69a92b8a6dbb7075b9520d35. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B MathVista Accuracy: 68.2. Result ID: 69a92b8a6dbb7075b9520d31. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### VLAA-Thinker-7B MathVista Accuracy: 68. Result ID: 69a92b8a6dbb7075b9520d34. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### R1-OneVision-7B MathVista Accuracy: 64.1. Result ID: 69a92b8a6dbb7075b9520d32. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7B"}. This record does not assert independent reproduction. ### GPT4o MathVista Accuracy: 61.4. Result ID: 69a92b8a6dbb7075b9520d2f. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Model Category":"Proprietary Model"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.