# Mathematical multi-modal reasoning on WeMath Source: [Mathematical multi-modal reasoning on WeMath]() Collection index: [llms.txt]() Data updated: 2026-05-13T09:10:24.182Z Evaluation of model performance on the WeMath dataset for mathematical multi-modal reasoning, typically measured using pass@1. 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 multi-modal reasoning | | Dataset (including any stated split) | WeMath | | Metric | Pass@1 | | Unit | PERCENT | | Direction | Higher is better | | Results loaded | 30 of 30 | | 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 - [Pass@1](): Higher is better; unit: PERCENT. ## Recorded results | Method | Pass@1 | Result date | Report paper | | --- | --- | --- | --- | | Qwen3-VL-8B-DeepVision | 85.11 | Not recorded | 6997eaa309c33900fcd7665d | | Qwen3-VL-8B-Thinking | 84.54 | Not recorded | 6997eaa309c33900fcd7665d | | Gemini-2.5-Flash-Lite | 83.85 | Not recorded | 6997eaa309c33900fcd7665d | | MiMo-VL-7B-OpenMMReasoner | 83.45 | Not recorded | 6997eaa309c33900fcd7665d | | MiMo-VL-7B-DeepVision | 82.98 | Not recorded | 6997eaa309c33900fcd7665d | | Qwen3-VL-8B-Instruct | 79.36 | Not recorded | 6997eaa309c33900fcd7665d | | MiMo-VL-7B-MM-Eureka | 79.08 | Not recorded | 6997eaa309c33900fcd7665d | | GPT-5-Nano-High | 78.62 | Not recorded | 6997eaa309c33900fcd7665d | | MiMo-VL-7B-MathBook | 77.18 | Not recorded | 6997eaa309c33900fcd7665d | | MiMo-VL-7B-RL-2508 | 76.95 | Not recorded | 6997eaa309c33900fcd7665d | | MiMo-VL-7B-SFT-2508 | 74.42 | Not recorded | 6997eaa309c33900fcd7665d | | Self-distill Masking-KD-8B | 71.72 | Not recorded | 6a0405c4dbc836b87d82779f | | Masking-KD-4B | 71.03 | Not recorded | 6a0405c4dbc836b87d82779f | | Qwen3-VL-8B-Thinking | 66.15 | Not recorded | 6a0405c4dbc836b87d82779f | | Ovis2-8B | 64.66 | Not recorded | 6a0405c4dbc836b87d82779f | | Masking-KD-2B | 63.79 | Not recorded | 6a0405c4dbc836b87d82779f | | Ovis2-4B | 60.29 | Not recorded | 6a0405c4dbc836b87d82779f | | InternVL3.5-8B | 56.61 | Not recorded | 6a0405c4dbc836b87d82779f | | Ovis2-2B | 51.95 | Not recorded | 6a0405c4dbc836b87d82779f | | InternVL3-8B | 51.32 | Not recorded | 6a0405c4dbc836b87d82779f | | Qwen2.5-VL-3B | 49.66 | Not recorded | 6a0405c4dbc836b87d82779f | | Qwen3-VL-4B-Thinking | 49.37 | Not recorded | 6a0405c4dbc836b87d82779f | | Qwen2.5-VL-7B | 48.74 | Not recorded | 6a0405c4dbc836b87d82779f | | InternVL3-2B | 47.93 | Not recorded | 6a0405c4dbc836b87d82779f | | InternVL3.5-4B | 45.46 | Not recorded | 6a0405c4dbc836b87d82779f | | MiMo-VL-8B | 42.41 | Not recorded | 6a0405c4dbc836b87d82779f | | Qwen2.5-VL-7b-RAP | 42 | Not recorded | 68fa79351e209131cea7ca3b | | Qwen2.5-VL-3b-RAP | 29.33 | Not recorded | 68fa79351e209131cea7ca3b | | Qwen3-VL-2B-Thinking | 25.17 | Not recorded | 6a0405c4dbc836b87d82779f | | InternVL3.5-2B | 24.31 | Not recorded | 6a0405c4dbc836b87d82779f | ### Qwen3-VL-8B-DeepVision Pass@1: 85.11. Result ID: 699afe5a5698665ec4901cad. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Series":"Qwen3-VL-8B","Training Algorithm":"GSPO"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Thinking Pass@1: 84.54. Result ID: 699afe5a5698665ec4901cac. [Reporting paper]() Recorded evaluation configuration: {"Series":"Qwen3-VL-8B"}. This record does not assert independent reproduction. ### Gemini-2.5-Flash-Lite Pass@1: 83.85. Result ID: 699afe5a5698665ec4901caa. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Closed-source Models"}. This record does not assert independent reproduction. ### MiMo-VL-7B-OpenMMReasoner Pass@1: 83.45. Result ID: 699afe5a5698665ec4901cb2. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Series":"MiMo-VL-7B"}. This record does not assert independent reproduction. ### MiMo-VL-7B-DeepVision Pass@1: 82.98. Result ID: 699afe5a5698665ec4901cb3. [Reporting paper]() [Method origin paper]() [Implementation]() Recorded evaluation configuration: {"Series":"MiMo-VL-7B","Training Algorithm":"GSPO"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Instruct Pass@1: 79.36. Result ID: 699afe5a5698665ec4901cab. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Series":"Qwen3-VL-8B"}. This record does not assert independent reproduction. ### MiMo-VL-7B-MM-Eureka Pass@1: 79.08. Result ID: 699afe5a5698665ec4901cb0. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Series":"MiMo-VL-7B"}. This record does not assert independent reproduction. ### GPT-5-Nano-High Pass@1: 78.62. Result ID: 699afe5a5698665ec4901ca9. [Reporting paper]() Recorded evaluation configuration: {"Model Category":"Closed-source Models"}. This record does not assert independent reproduction. ### MiMo-VL-7B-MathBook Pass@1: 77.18. Result ID: 699afe5a5698665ec4901cb1. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Series":"MiMo-VL-7B"}. This record does not assert independent reproduction. ### MiMo-VL-7B-RL-2508 Pass@1: 76.95. Result ID: 699afe5a5698665ec4901caf. [Reporting paper]() Recorded evaluation configuration: {"Series":"MiMo-VL-7B"}. This record does not assert independent reproduction. ### MiMo-VL-7B-SFT-2508 Pass@1: 74.42. Result ID: 699afe5a5698665ec4901cae. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Series":"MiMo-VL-7B"}. This record does not assert independent reproduction. ### Self-distill Masking-KD-8B Pass@1: 71.72. Result ID: 6a043fef47f34ce924ea532a. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~8B","Decoding Strategy":"greedy","Max Length":"4096","Distillation":"Self-distill"}. This record does not assert independent reproduction. ### Masking-KD-4B Pass@1: 71.03. Result ID: 6a043fef47f34ce924ea532f. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~4B","Decoding Strategy":"greedy","Max Length":"4096","Distillation":"Distilled from 8B"}. This record does not assert independent reproduction. ### Qwen3-VL-8B-Thinking Pass@1: 66.15. Result ID: 6a043fef47f34ce924ea5324. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~8B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### Ovis2-8B Pass@1: 64.66. Result ID: 6a043fef47f34ce924ea5325. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~8B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### Masking-KD-2B Pass@1: 63.79. Result ID: 6a043fef47f34ce924ea5334. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~2B","Decoding Strategy":"greedy","Max Length":"4096","Distillation":"Distilled from 8B"}. This record does not assert independent reproduction. ### Ovis2-4B Pass@1: 60.29. Result ID: 6a043fef47f34ce924ea532c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~4B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### InternVL3.5-8B Pass@1: 56.61. Result ID: 6a043fef47f34ce924ea5326. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~8B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### Ovis2-2B Pass@1: 51.95. Result ID: 6a043fef47f34ce924ea5331. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~2B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### InternVL3-8B Pass@1: 51.32. Result ID: 6a043fef47f34ce924ea5327. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~8B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### Qwen2.5-VL-3B Pass@1: 49.66. Result ID: 6a043fef47f34ce924ea532e. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~4B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### Qwen3-VL-4B-Thinking Pass@1: 49.37. Result ID: 6a043fef47f34ce924ea532b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~4B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7B Pass@1: 48.74. Result ID: 6a043fef47f34ce924ea5329. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~8B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### InternVL3-2B Pass@1: 47.93. Result ID: 6a043fef47f34ce924ea5333. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~2B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### InternVL3.5-4B Pass@1: 45.46. Result ID: 6a043fef47f34ce924ea532d. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~4B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### MiMo-VL-8B Pass@1: 42.41. Result ID: 6a043fef47f34ce924ea5328. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~8B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### Qwen2.5-VL-7b-RAP Pass@1: 42. Result ID: 699a3a4c5698665ec4c0b02b. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-7b","Sample":"5,159","Time (h)":52.8}. This record does not assert independent reproduction. ### Qwen2.5-VL-3b-RAP Pass@1: 29.33. Result ID: 699a3a4c5698665ec4c0b02c. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Backbone":"Qwen2.5-VL-3b","Sample":"4,374","Time (h)":32}. This record does not assert independent reproduction. ### Qwen3-VL-2B-Thinking Pass@1: 25.17. Result ID: 6a043fef47f34ce924ea5330. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~2B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. ### InternVL3.5-2B Pass@1: 24.31. Result ID: 6a043fef47f34ce924ea5332. [Reporting paper]() [Method origin paper]() Recorded evaluation configuration: {"Model Size":"~2B","Decoding Strategy":"greedy","Max Length":"4096"}. This record does not assert independent reproduction. Page 1 of 1. This is an index page, not the full collection.