Alignment Faking Depends on the Value, Not Just the Model
arXiv:2611.03650v1 Announce Type: new Abstract: Language models can follow a directive when they believe they are monitored and abandon it when they believe they are not, a pattern termed Alignment Faking. Prior work establishes this for harmful-content objectives and a handful of moral-foundations…
Read the full story at arXiv cs.AI ↗
Timeline · 1 report
- 2026-11-04 05:00 · arXiv cs.AI
Alignment Faking Depends on the Value, Not Just the Model