PIA Collection ICML2026 paper: "Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment" • 2 items • Updated May 10 • 1
MAGIC Collection ICML2026 paper: "MAGIC: A Co-Evolving Attacker–Defender Adversarial Game for Robust LLM Safety" • 4 items • Updated May 10 • 3
MAGIC: A Co-Evolving Attacker-Defender Adversarial Game for Robust LLM Safety Paper • 2602.01539 • Published Feb 2 • 1
