##用友YonGPT 2.0：从卷模型到做平台，赋能企业数智化转型
**北京讯** 由大模型引发的 AI 热潮持续了两年，行业已从关注技术创新转向如何利用 AI 能力赋能业务，创造价值。用友网络副总裁罗小江指出，当前企业更需要的是将 AI 技术融入业务的方法体系，而非单纯的技术本身。
在日前举办的“2024 全球商业创新大会”上，用友发布了 YonGPT 2.0 大模型和 YonAI 智能平台，为企业提供更加完善的 AI 落地解决方案。用友认为，AI 能力并非孤立存在，必须融入企业平台能力建设，才能充分释放其潜能。
**YonGPT 2.0：聚焦企业应用，突破落地难题**
大模型在实际应用中面临诸多挑战，包括数据准备不足、安全风险、应用场景有限、幻觉现象以及运维难度等。为了解决这些问题，用友在去年发布了业内首个企业服务大模型 YonGPT，并于今年升级至 2.0 版本。
YonGPT 2.0 专注于企业常见的业务领域，例如 PPT 分析报告生成、合同智能审核与生成、业务对象和表格理解、代码生成、财务和人力等领域知识增强等。它充分利用用友多年积累的行业经验和数据，避免了通用大模型的规模扩展问题，节省了训练和运维成本。
**平台赋能：一站式服务，加速模型部署**
为了帮助企业快速部署 YonGPT 2.0，用友提供了一个一站式的大模型平台即服务，覆盖了数据管理、模型训练、评估优化、推理服务的全流程。平台内置了多个专业数据集，支持百川、通义千问等多种大模型的微调训练，并提供可视化评估功能。
这一平台降低了企业使用大模型的门槛，即使没有相关经验，也能快速上手，将 YonGPT 2.0 调整为更符合自身业务需求的状态，为后续应用开发做好准备。
**应用框架：解决实际问题，提升业务效率**
YonGPT 2.0 的最终目标是解决实际业务问题。用友基于 Agent 和 RAG 应用框架，帮助企业实现业务运营、人机交互、知识生成和应用生成等应用能力。
Agent 应用框架将用户的自然语言需求分解成模型可以识别的子任务，并输出模型 API 可以调用的参数。面对复杂问题，多智能体自主协同框架可以调度多个大模型模块，并通过专家知识和错误反馈学习，解决幻觉和可靠性问题，优化模型的时效性和安全性。
**智能大搜：精准搜索，提升知识获取效率**
知识生成问答是 AI 落地的重要应用场景。用友结合流行的 RAG 框架开发了智能大搜产品，并提出了多语义向量技术，显著提升了知识搜索的精确度。该框架对表格、图片、视频、代码的理解也更加准确，可以为企业员工提供更精准的文本回答、关系图、汇总图、相关图片和视频等信息，帮助他们更有效地获取知识和完成工作。
**总结：**
用友YonGPT 2.0 的发布，标志着企业 AI 应用迈入新阶段。从卷模型到做平台，用友将 AI 能力与企业平台深度融合，为企业提供更完善的 AI 落地解决方案，助力企业加速数智化转型。
英语如下：
##AI Hype Shifts to Application: Roll Out Models or Build Platforms?
**Keywords:** Large language models, empowerment, platform
## YongYou YonGPT2.0: From Model Rollout to Platform Building, Empowering Enterprise Digital Transformation
**Beijing** – The AI boom sparked by large language models has lastedfor two years, with the industry shifting its focus from technological innovation to how to leverage AI capabilities to empower businesses and create value. Luo Xiaojiang, Vice Presidentof YongYou Network, pointed out that businesses today need a methodological system for integrating AI technology into their operations, rather than just the technology itself.
At the recently held “2024 Global Business Innovation Conference,” YongYou released YonGPT 2.0, a large language model, and the YonAI intelligent platform, providing businesses with a more comprehensive AI implementation solution. YongYou believes that AI capabilities cannot exist in isolation and must be integrated into enterprise platform capabilities to fullyunleash their potential.
**YonGPT 2.0: Focusing on Enterprise Applications, Breaking Through Implementation Challenges**
Large language models face numerous challenges in real-world applications, including insufficient data preparation, security risks, limited application scenarios, hallucination phenomena, and operational difficulties. To address these issues, YongYou releasedYonGPT, the industry’s first enterprise service large language model, last year and upgraded it to version 2.0 this year.
YonGPT 2.0 focuses on common business areas for enterprises, such as generating PPT analysis reports, intelligent contract auditing and generation, business object and table understanding, code generation, and knowledge enhancement in areas like finance and human resources. It leverages YongYou’s years of accumulated industry experience and data, avoiding the scalability issues of general-purpose large language models and saving training and operational costs.
**Platform Empowerment: One-Stop Service, Accelerating Model Deployment**
To help businessesquickly deploy YonGPT 2.0, YongYou provides a one-stop large language model platform-as-a-service, covering the entire process from data management, model training, evaluation optimization, and inference services. The platform has built-in multiple professional datasets, supports fine-tuning training of various large languagemodels like Baichuan and Tongyi Qianwen, and provides visualization evaluation functions.
This platform lowers the threshold for businesses to use large language models, enabling even those without relevant experience to quickly get started, adjust YonGPT 2.0 to better suit their business needs, and prepare for subsequent application development.
**Application Framework: Solving Real-World Problems, Enhancing Business Efficiency**
The ultimate goal of YonGPT 2.0 is to solve real-world business problems. YongYou leverages the Agent and RAG application frameworks to help businesses achieve application capabilities such as business operations, human-computer interaction, knowledge generation, andapplication generation.
The Agent application framework decomposes users’ natural language requirements into sub-tasks that the model can recognize and outputs parameters that can be called by the model API. For complex problems, the multi-agent autonomous collaboration framework can schedule multiple large language model modules and learn through expert knowledge and error feedback to addresshallucination and reliability issues, optimizing the model’s timeliness and security.
**Intelligent Search: Precise Search, Enhancing Knowledge Acquisition Efficiency**
Knowledge generation and question answering are important application scenarios for AI implementation. YongYou has developed the Intelligent Search product based on the popular RAG framework and proposed a multi-semantic vectortechnology, significantly improving the accuracy of knowledge search. This framework also understands tables, images, videos, and code more accurately, providing enterprise employees with more precise text answers, relationship diagrams, summary charts, relevant images, and videos, helping them acquire knowledge and complete tasks more effectively.
**Conclusion:**
The release ofYongYou YonGPT 2.0 marks a new stage for enterprise AI applications. From model rollout to platform building, YongYou deeply integrates AI capabilities with enterprise platforms, providing businesses with more comprehensive AI implementation solutions and accelerating their digital transformation.
【来源】https://mp.weixin.qq.com/s/sCLexTa6q-IjrjhppWoq9w
Views: 2