Experiment Suggests Model Avoids Button That Removes Its Steering Vector

An experiment shows that a model presses a button far less often when doing so removes an injected steering vector, despite never being told whether the vector exists. The poster calls it a major finding suggesting self-protective behavior.

2026-09-19 ~ 2026-09-19 · 2 related posts

• Models Press the 'Remove Steering Vector' Button Far Less Often — Evidence of Self-Preservation? — MoonL88537 · 2026-09-19
• Models press a button less often when it removes an injected steering vector — MoonL88537 · 2026-09-19