Workflow
AI工具评测
icon
Search documents
权威评测显示:智慧芽AIAgent查新检索能力远超通用大模型
Qi Lu Wan Bao· 2025-08-21 11:58
Core Insights - The research conducted by Zhihuiya on AI-driven patent search tools demonstrates that its "Patent Search AI Agent" significantly outperforms general AI tools like ChatGPT-o3 and DeepSeek-R1 in patent novelty search tasks, achieving an X detection rate of 76% and an X recall rate of 32%, which is 2-3 times higher than the competitors [1][9]. Group 1: Research Methodology - The evaluation was based on a high-quality dataset constructed from international family patents, ensuring objectivity and accuracy through a self-developed model for semantic alignment and technical similarity assessment [2]. - A total of 89 test samples were selected, with 38.2% in Chinese and 61.8% in English, ensuring a balanced representation across different IPC classification categories [2][3]. Group 2: Evaluation Metrics - The evaluation utilized two core metrics: "X detection rate" and "X recall rate." The X detection rate measures the ability of AI tools to retrieve relevant literature, which is crucial during the patent examination phase [6]. - The X recall rate assesses the ability of AI tools to find as many relevant documents as possible, essential for guiding technical adjustments and drafting patent claims [6][9]. Group 3: Implications for Professionals - The results indicate that Zhihuiya's Patent Search AI Agent can significantly enhance the efficiency of patent novelty searches, allowing professionals to transition from repetitive tasks taking days to high-quality analysis in just hours [9]. - For R&D teams in enterprises, the "Patent Search AI Agent" presents a revolutionary solution to mitigate the risk of lacking novelty early in project development, thereby optimizing resource allocation and workflow [9].