Researchers Debate Whether RL Makes Unlearning Easy
A discussion argues that RL-era unlearning is nearly trivial—just exclude relevant RL environments from training—though critics note that offensive cyber capabilities are harder to separate, since finding CVEs favors defense.
2026-11-13 ~ 2026-11-13 · 3 related posts
- Unlearning Gets Trivial in the RL Era: Just Exclude the RL Envs — maksym_andr · 2026-11-13
- Debate: RL makes unlearning trivial, but cyber offense-defense can be separated — xeophon · 2026-11-13
- AI Safety Debate: Are a Model's Offensive and Defensive Cyber Skills Really '~the Same'? — maksym_andr · 2026-11-13