GenExam is a benchmark dataset for evaluating expert-level intelligence in integrated understanding, reasoning, and generation tasks. The dataset, created by OpenGVLab, focuses on multidisciplinary text-to-image exams. It was last updated on October 6, 2025.
Use Cases
- Benchmarking AI model performance on multidisciplinary exams based on the described exam-style tasks
- Evaluating integrated understanding and reasoning capabilities in multimodal systems based on the dataset's focus
- Testing text-to-image generation quality against expert-level intelligence benchmarks
Strengths
- Dataset is explicitly designed for expert-level intelligence evaluation, focusing on integrated understanding, reasoning, and generation
- The benchmark is multidisciplinary, covering multiple domains within its exam structure
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
- OpenGVLab
- Freshness
- Last updated 2025-10-06 19:13:40