LLM and VLM Workflow for RISC-V Supply Chain Analysis
An academic paper has introduced a workflow that utilizes LLMs to analyze RISC-V supply chains, combining Vision-Language Models (VLMs) with Model-Driven Engineering (MDE). This method tackles the challenges posed by diverse and unstructured supply chain data, employing LLMs for text comprehension and VLMs for extracting insights from visual materials such as diagrams, tables, and scanned documents. The models pinpoint essential entities and their connections, which are structured into a knowledge graph that illustrates supply chain elements and their interrelations. MDE strategies and constraint-based modeling facilitate the formal validation of dependencies, identification of bottlenecks, and risk evaluation. This integration of semantic insights from LLMs/VLMs with MDE's formal analysis yields thorough, multimodal data-driven insights. The paper is available on arXiv with the identifier 2605.15223.
Key facts
- Paper presents an LLM-empowered workflow for RISC-V supply chain analysis
- Integrates Vision-Language Models (VLMs) and Model-Driven Engineering (MDE)
- Addresses heterogeneous and unstructured supply chain data
- LLMs used for textual understanding, VLMs for visual artifacts
- Identifies entities and relationships organized into a knowledge graph
- MDE enables formal validation, bottleneck detection, risk assessment
- Published on arXiv with identifier 2605.15223
- Approach supports comprehensive multimodal data-driven insights
Entities
Institutions
- arXiv