Defending against Adaptive Prompt Injection Attacks via Reasoning-enabled Task Alignment

Feeds arXiv
1
This is a companion discussion topic for the original entry at https://arxiv.org/abs/2606.15441