JARVIS: Advancing AGI Research Through a Collaborative System JARVIS is dedicated to advancing the exploration of Artificial General Intelligence (AGI) by delivering cutting-edge research to the broader community. Our system operates as a collaborative framework, integrating a large language model (LLM) as the central controller and multiple specialized expert models sourced from the Hugging Face Hub as execution modules. The system follows a structured four-stage workflow:
Task Planning: The LLM (ChatGPT) analyzes user requests, interprets their intent, and decomposes them into distinct, manageable tasks. Model Selection: Based on the planned tasks, the LLM selects appropriate expert models from the Hugging Face Hub by evaluating their descriptions and capabilities. Task Execution: The selected expert models are invoked to process their respective tasks, with results aggregated and returned to the LLM. Response Generation: The LLM synthesizes outputs from all expert models, integrating their predictions into a coherent and comprehensive response. This collaborative approach enhances task efficiency, leveraging the strengths of multiple AI models to produce refined, high-quality responses.