Steering vectors align LLMs with human values

Automated excerpt

Linear directions extracted from large‑language‑model activation distributions map onto human‑value axes with measurable fidelity. The study demonstrates that these steering vectors preserve the full geometry of a theory‑driven value space, not merely isolated behavioral tweaks. This contrast indicates that shortcut‑based steering can hit target scores without embedding the intended value relationships, underscoring the uniqueness of linear, distribution‑derived vectors.

Selected automatically from source text; not independently written or fact-checked. Read the original for full context.