U1-OCR号联合音响发射U1-OCR号是第一种工业AI号模型,用于提高文件理解程度,提高准确性和真实用途的推理。 Unisound launches U1-OCR, the first industrial AI model for advanced document understanding, boosting accuracy and reasoning for real-world use.

U1-OCR是全世界第一个用于文件情报的工业级基础模型,它标志着OCR 3.0的启动。 Unisound has launched U1-OCR, the world’s first industrial-grade foundation model for document intelligence, marking the start of OCR 3.0. AI模型提供深入的语义理解,使自动分类和精确的信息从复杂、无组织或低质量文件中提取成为可能。 The AI model delivers deep semantic understanding, enabling automatic classification and precise information extraction from complex, disorganized, or low-quality documents. 它使用语义驱动的动态焦点战略和先进技术,如多肯预测和加强学习,以提高推理效率80%以上,并实现OmniDocBench V1.5的95.1。 It uses a semantic-driven, dynamic focus strategy and advanced techniques like Multi-Token Prediction and reinforcement learning to improve reasoning efficiency by over 80% and achieve 95.1 on OmniDocBench V1.5. 它设计用于现实世界,支持私人的预先部署,确保数据隐私,并保持50+类文件的高度准确性-99以上,甚至含模糊或多语种内容。 Designed for real-world use, it supports private on-premise deployment, ensures data privacy, and maintains high accuracy across 50+ document types—over 99%—even with blurry or multilingual content. 其协调文本-文字-语义结构使完全可追踪性、审计准备状态和零点逻辑核查成为可能,将AI从文本识别推进到认知推理。 Its coordinate-text-semantics architecture enables full traceability, audit readiness, and zero-shot logic verification, advancing AI from text recognition to cognitive reasoning.