OCR-Reasoning Benchmark: Unveiling the True Capabilities of MLLMs in Complex Text-Rich Image Reasoning 文章

ArXiv CS.CV2026-05-27NEWSen作者: Mingxin Huang, Yongxin Shi, Dezhi Peng, Songxuan Lai, Zecheng Xie, Lianwen Jin
大语言模型
查看原文 →
关系图谱

OCR-Reasoning Benchmark: Unveiling the True Capabilities of MLLMs in Complex Text-Rich Image Reasoning · 相关技术

相关技术

iOSLarge language modelSSELLMORMMultimodal Large Language Models (MLLMs)ODEPLAlanguage model多模态Straight-Through EstimatorStanSPAReferring expression comprehension (REC)Parts-of-Speech (POS) tagsOTAOCRNLILMMHISGranular Alignment ParadigmForEffort Metric AttentionENADiTARGANN