We propose ‘model integrity’ as an overlooked challenge in aligning LLM agents.

In this paper, we clarify what’s meant by human values, and how we can align AI to them.

Beyond Constitutional AI; Our first trial with 500 Americans; How democratic processes can generate an LLM we can trust.

An alternative to Constitutional AI or simple RLHF-based approaches for fine-tuning LLMs based on moral information from diverse populations.