NumBench
Explore score history ↗Evaluates counting accuracy in text-to-image models using 640,000 prompts across 1,600 categories and counts from 1 to 100, with factorial variation of composition, layout, and appearance.
- Released
- 2026-08-28
- Institutions
- Indian Institute of Technology Jodhpur · Shiv Nadar University
- Evaluates
- General AI, Vision & 3D, cs.CV
- Openness
- unknown
- Importer
- Claire Radar
- Review status
- ai-reviewed
NumBench paper
No reported scores are on record for this benchmark yet.