China News Service, Hangzhou, September 28 (Bao Mengni) On September 28, at the 2026 Hangzhou International Talent Exchange and Project Cooperation Conference, the Hangzhou Technology Transfer and Transformation Center (hereinafter referred to as the "Hangzhou Transfer Center") officially released version 3.0 of the Wise Man Large Model. This is the third major upgrade of the platform after its debut as the first large-scale achievement transformation model in China in 2024 and the launch of version 2.0 in 2025.
The wise model is defined as a "AI technical manager who is online 24/7". It builds an industry knowledge base through the fusion of multi-source heterogeneous data, combines domain semantic retrieval, knowledge correlation analysis and large model reasoning, and is committed to solving the three long-standing industry pain points in the field of scientific and technological achievement transformation: difficulty in gathering corporate demand, difficulty in evaluating the value of achievements, and difficulty in accurately connecting supply and demand, and breaking through information barriers between universities, colleges, proof-of-concept centers, pilot platforms and technology companies.
According to the relevant person in charge of the Hangzhou Transfer Center, the core capabilities of the Wise Man Model 3.0 version can be summarized in three keywords: broad, deep and accurate.
"Extensive" is a comprehensive leap in the data base. Version 3.0 of the Wise Large Model has realized the leap from Zhejiang Province data to nationwide data. Through multi-source heterogeneous data fusion, entity disambiguation and cross-source alignment, the Hangzhou Transfer Center associates and integrates information on enterprises, experts and results scattered in different channels, builds a high-quality exclusive database, and creates a tens of billions of data base. At present, on the enterprise side, the model has stored information on 2.2 million technology companies across the country; on the university side, it has built a multi-dimensional scientific research capability portrait of 153,000 experts and scholars around research directions, representative achievements and technical expertise, covering more than 1,100 universities and colleges. On the achievement side, more than 1.3 million scientific and technological achievements have been standardized and stored in the database, and a high-quality data set covering 1.18 billion patents, papers, recruitment, projects and other information has been constructed, completely breaking the information barriers between regions, schools, enterprises and industries.
"Deep" refers to the improvement of penetration of demand analysis. The technical requirements put forward by enterprises are often general and vague. Version 3.0 of the Wise Man Model can dismantle the vague technical problems of enterprises layer by layer and match them one by one. By integrating multiple rounds of reasoning capabilities, the model transforms technical bottlenecks into retrievable and matchable technical requirements, realizing the intelligentization of the entire process from "problem analysis" to "requirements modeling" to "outcome matching", and predicting multi-step technical research paths like playing chess.
"Accurate" refers to the exponential improvement in the accuracy of matching supply and demand. In the research and development of version 3.0 of the Wise Big Model, the R&D team combined vector semantic retrieval and knowledge graph correlation analysis to align the technical semantics of enterprise needs and scientific and technological achievements, identify potential technical correlations behind different expressions, cleans more than 6 billion pieces of multi-source data, integrate enterprise, achievement, expert and other information into a unified knowledge graph, and has gone through more than 200,000 iterations of matching and optimization. After testing, the accuracy of enterprise technology prediction is close to 80%. (over)
