AI Dynamics

Global AI News Aggregator

About

LLM domain training should produce manuals not weight updates

yes exactly, "training" an LLM on a target domain should output a manual not a weight diff.

→ View original post on X — @karpathy