introducing @relignai
an open-source reinforcement learning framework designed to help AI agents get better at completing complex tasks.
currently agents are auto-regressive, meaning their success rate diminishes with every consecutive step. if the success rate of each step is
roadmap__
phase i - launch__complete
phase ii - develop a strong reasoning library | grpo, full docs, whitepaper, modularization, expand reasoning stack, expand dev community
phase iii - scaling up | cloud runner, 70B+ parameter reasoning model, relign toolkit, research
relign.gitbook.io/relign
relign gitbook v1 live - covering concept, vision, mission, roadmap, use cases
notable quote: relign is a step towards artificial general intelligence. the technology has far-reaching application, because it is a fundamental tool that teaches base models
