ScienceQA
by Lu et al. / UCLA · free · Last verified 2026-03-17
ScienceQA is a large-scale multimodal benchmark featuring 21,208 science questions for grades 3-12. It uniquely combines visual diagrams and textual contexts, requiring models to perform complex reasoning. Each question includes multiple-choice options, a detailed lecture, and a step-by-step explanation for the correct answer.
https://scienceqa.github.io ↗B
B—Above Average
Adoption: B+Quality: AFreshness: B+Citations: AEngagement: F
Specifications
- License
- CC BY-NC-SA 4.0
- Pricing
- free
- Capabilities
- multimodal question answering, visual reasoning and diagram understanding, scientific knowledge retrieval, natural language understanding, chain-of-thought reasoning evaluation, explanation generation, evaluating models on K-12 science topics
- Integrations
- Use Cases
- [object Object], [object Object], [object Object], [object Object]
- API Available
- No
- Evaluated Models
- gpt-4o, claude-opus-4, gemini-2-5-pro, llava-1.6
- Metrics
- accuracy
- Methodology
- 21,208 multiple-choice questions across 26 topics. Models answer with or without context images; performance averaged across image/text-only subsets. Chain-of-thought explanations optionally evaluated for rationale quality.
- Last Run
- 2026-02-02
- Tags
- benchmark, science-qa, multimodal-reasoning, visual-question-answering, vqa, k12-education, chain-of-thought, explanation-generation, natural-language-processing, evaluation
- Added
- 2026-03-17
- Completeness
- 0.7%
Index Score
68Adoption
76
Quality
88
Freshness
73
Citations
80
Engagement
0