What AI says about MARTI

MARTI is an open-source framework for training LLM-based multi-agent systems (MAS) with reinforcement learning, enabling centralized multi-agent interactions with distributed policy training. It supports built-in graph-based workflows and compatibility with popular third-party multi-agent frameworks to enable scalable, adaptive RL pipelines. The project aims to advance reasoning and collective intelligence beyond individual LLMs and is in an early experimental stage, welcoming collaboration.