SciDocBench
Explore score history ↗SciDocBench evaluates scientific document understanding with 124 expert-authored questions across seven capability groups and 19 subtasks in five scientific domains. Each question is tested under four conditions (English vs Chinese and all-images-first vs interleaved), yielding 496 instances. Performance is reported on a 100-point scale. Data and pipeline are released via a public repository.
- Released
- 2026-09-04
- Institutions
- Chinese University of Hong Kong · Shanghai Artificial Intelligence Laboratory · New York University
- Evaluates
- General AI, Language & Knowledge, cs.AI
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- ai-name-audit-deferred
SciDocBench paper and code
No reported scores are on record for this benchmark yet.