📖 The AI Tool Bible
safety

Alignment

The practice of shaping a model's behaviour to match human intent and values — via RLHF, DPO, constitutional AI, or filtered training data.

Related terms