Instead of manually deciding how to fine-tune an LLM, TREX uses AI agents to automatically explore training strategies, learn from past experiments, and optimize performance—treating the entire fine-tuning process as a searchable problem.
TREX is a multi-agent system that automates the entire process of fine-tuning large language models, from analyzing requirements to training and evaluation. It uses a tree-based search approach to explore different training strategies efficiently, reusing past results and learning from experiments.