# Document OCR Optimization

The Perception Encoder (PE) is an advanced image and video understanding encoder trained through vision-language learning, achieving state-of-the-art performance on various visual tasks.