2,000 vision-text multimodal physics problems spanning from middle school to doctoral qualification exams. The SeePhys benchmark systematically evaluates LLMs/MLLMs on tasks integrating complex scientific diagrams with theoretical derivations. It was created by SeePhys and last updated on 2025-10-28.
Use Cases
- Benchmarking AI physics reasoning based on multimodal problems
- Evaluating model performance on scientific diagram interpretation
- Assessing AI on educational assessment tasks across academic levels
Strengths
- Contains 2,000 distinct physics problems
- Covers a systematic range from middle school to doctoral qualification exams
Limitations
- Column-level documentation is absent; field semantics must be inferred after download
- Row count is unknown, which may limit suitability assessment
Provenance
- Source
- SeePhys
- Freshness
- Last updated 2025-10-28 03:26:37