Measuring Multimodal Mathematical Reasoning with MATH-Vision Dataset
This provides a more comprehensive benchmark for researchers developing multimodal AI systems for mathematical reasoning, though it is incremental as it extends existing evaluation datasets.
The authors tackled the limited diversity in existing benchmarks for evaluating multimodal mathematical reasoning by creating the MATH-Vision dataset with 3,040 problems from real competitions across 16 disciplines and 5 difficulty levels, revealing a significant performance gap between current large multimodal models and humans.
Recent advancements in Large Multimodal Models (LMMs) have shown promising results in mathematical reasoning within visual contexts, with models approaching human-level performance on existing benchmarks such as MathVista. However, we observe significant limitations in the diversity of questions and breadth of subjects covered by these benchmarks. To address this issue, we present the MATH-Vision (MATH-V) dataset, a meticulously curated collection of 3,040 high-quality mathematical problems with visual contexts sourced from real math competitions. Spanning 16 distinct mathematical disciplines and graded across 5 levels of difficulty, our dataset provides a comprehensive and diverse set of challenges for evaluating the mathematical reasoning abilities of LMMs. Through extensive experimentation, we unveil a notable performance gap between current LMMs and human performance on MATH-V, underscoring the imperative for further advancements in LMMs. Moreover, our detailed categorization allows for a thorough error analysis of LMMs, offering valuable insights to guide future research and development. The project is available at https://mathvision-cuhk.github.io