OCRBench v2

Measures: 10,000 human-verified English and Chinese QA pairs over 31 scenarios (scene text, receipts, formulas, diagrams, documents) testing text recognition, localization and reasoning; per-task scores averaged to 0-100, with a private 1,500-image test set for the leaderboard.

Maintainer: Yuliang Liu et al., Huazhong University of Science and Technology
Introduced: 2024
Status: Active
Saturation: Headroom
Typical provenance: Independent
In scoring set: Yes

Paper Official leaderboard

Leaderboard

Show independent results only

Every result we hold, whoever ran it. Only independent and third-party rows count towards a score.

No results to show.