A multimodal mathematics dataset collected from real middle school exams in China, featuring open-ended problems. It was created by THU-KEG and last updated on June 30, 2024. The dataset is annotated with fine-grained three-dimensional labels for difficulty, grade, and knowledge points.
Use Cases
- Benchmarking multimodal model performance on open-ended math problems based on the dataset's core purpose.
- Analyzing problem difficulty based on the difficulty labels derived from student exam scores.
- Studying the distribution of mathematical concepts based on the annotated knowledge points.
- Evaluating grade-level appropriateness of AI-generated solutions based on the provided grade labels.
Strengths
- Data is sourced from real middle school exams in China, providing authentic assessment material.
- Problems are annotated with three-dimensional labels: difficulty, grade, and knowledge points.
- Difficulty levels are determined based on average student exam scores.
Limitations
- Column-level documentation is absent; field semantics must be inferred after download.
- Row count is unknown, which may limit suitability assessment.
- Description metadata is limited; actual data quality requires manual inspection after download.
Provenance
- Source
- THU-KEG
- Collection Method
- Collected from real middle school exams in China.
- Freshness
- Last updated 2024-06-30 14:20:05; freshness should be verified.
- Geography
- China