NeurIPS Paper Makes Agent Defenses Evolvable as Skills

A NeurIPS-2026 paper proposes Defense-as-Skill, arguing agent defenses should evolve like skills instead of remaining static policies. Using MCTS, the method cuts an agent guard's iterative attack success rate from 0.30 to 0.078 over 9 rounds.

2026-10-14 ~ 2026-10-15 · 2 related posts

• Defense-as-Skill: NeurIPS paper makes agent runtime guards evolvable, with 206-case attack dataset — jiqizhixin · 2026-10-14

1 near-duplicate retellings: 青稞AI