Shifting the Gradient: Understanding How Defensive Training Methods Protect Language Model Integrity

arXiv:2604.16423 · cs.LG, cs.AI · Submitted 2026-04-03 · Read on arXiv

cs.LG, cs.AI

Submitted: 2026-04-03

Updated: 2026-11-07

Terminology

Sources

Related papers