Tags: #safety

A modular open-source framework for training constrained value-aligned Large Language Models (LLMs) using Safe Reinforcement Learning from Human Feedback (RLHF).

Provides safe and friendly Rust bindings to various *nix platform APIs, offering a robust alternative to the unsafe `libc` crate.