Large-scale language model twin system
The large-scale language model twin system addresses customization and security challenges by integrating a user interface, digital twin, and multiple LLM agents for secure and accurate responses in factory environments.
Patent Information
- Authority / Receiving Office
- JP · JP
- Patent Type
- Patents
- Current Assignee / Owner
- MITSUBISHI ELECTRIC CORP
- Filing Date
- 2025-08-07
- Publication Date
- 2026-05-22
AI Technical Summary
Existing LLM-based agents for real-time data analysis and simulation in factories face challenges with customization, management complexity, and security concerns, leading to inaccurate responses and high-performance requirements that complicate deployment.
A large-scale language model twin system comprising a user interface, digital twin, and multiple LLM agents, including manager, motion, expert, and line agents, enables secure, tailored solutions for factories by integrating peer-to-peer encrypted communication and real-time data analysis.
The system allows for rapid construction of LLM solutions tailored to specific factories, ensuring secure and accurate responses through data analysis and simulation, simplifying management and enhancing response accuracy.
Smart Images
Figure 0007864275000001 
Figure 0007864275000002 
Figure 0007864275000003