Preprint
Aug 2026
WildHandBench: A Benchmark for Handwritten Text Understanding that Challenges MLLMs and Humans
A Prior-Driven Error (PDE) metric is introduced that quantifies whether errors originate from language priors rather than visual evidence rather than visual evidence in order to expose systematic reliance on language priors that conventional accuracy metrics cannot capture.
Jun Zhang, Qiao Zhao, Cheng Cui et al.
· 0 citations