Researchers have introduced Chronicles-OCR, a new benchmark designed to test the cross-temporal perception abilities of Vision Large Language Models (VLLMs) on Chinese characters. This benchmark covers the complete evolutionary trajectory of Chinese scripts, from ancient tortoise shells to modern calligraphy, addressing the lack of datasets that capture systematic visual shifts over thousands of years. Chronicles-OCR includes 2,800 balanced images and proposes a novel annotation paradigm to handle drastic morphological variations, offering four tasks to evaluate VLLMs' limitations in historical text perception. AI
Summary written by gemini-2.5-flash-lite from 1 source. How we write summaries →
IMPACT Provides a new evaluation tool for VLLMs to assess their robustness on historical scripts, potentially improving AI's utility in digital humanities.
RANK_REASON The cluster describes a new academic paper introducing a novel benchmark for evaluating AI models. [lever_c_demoted from research: ic=1 ai=1.0]