Show HN: Multi Agent Protocol for AI Scientist by Hexo Labs (github.com)

Hexo Labs has introduced the "Socrates" protocol, a novel multi-agent framework designed to enhance the capabilities of AI research agents by pairing a tool-using "Scientist" with a question-only advisor, known as Socrates. This setup requires each research plan to be approved by Socrates before execution, demonstrating significant performance improvements. In trials across five Kaggle competitions, the combined approach reportedly boosted test scores by an average of 55.9% compared to agents operating unassisted. Key to this success is Socrates' ability to engage the Scientist in stateful discussions, prompting targeted analyses and refined experimental strategies. The Socrates framework is particularly significant for the AI/ML community as it introduces a structured questioning mechanism that can unlock latent knowledge in AI agents, encouraging deeper exploration and better iterative results in complex problem domains. The implementation allows for both sequential and evolutionary approaches, with configurable flags enabling different operational modes, including a full Socratic review and a baseline control condition. This versatile architecture supports rapid experimentation and could lead to breakthroughs in how AI systems tackle research challenges.