PIA ICML2026 paper: "Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment" XiaoyuWen/PIA-Persona-Dataset Viewer • Updated May 10 • 35 • 33 XiaoyuWen/PIA Updated May 10
MAGIC ICML2026 paper: "MAGIC: A Co-Evolving Attacker–Defender Adversarial Game for Robust LLM Safety" XiaoyuWen/MAGIC-Qwen2.5-14B-Instruct Text Generation • Updated Feb 4 • 4 • 1 XiaoyuWen/MAGIC-Qwen2.5-7B-Instruct Text Generation • 8B • Updated Feb 4 • 17 • • 2 XiaoyuWen/MAGIC-Llama3.1-8B-Instruct Text Generation • Updated Feb 4 • 8 XiaoyuWen/MAGIC Text Generation • Updated Feb 4 • 1
PIA ICML2026 paper: "Disentangling Intent from Role: Adversarial Self-Play for Persona-Invariant Safety Alignment" XiaoyuWen/PIA-Persona-Dataset Viewer • Updated May 10 • 35 • 33 XiaoyuWen/PIA Updated May 10
MAGIC ICML2026 paper: "MAGIC: A Co-Evolving Attacker–Defender Adversarial Game for Robust LLM Safety" XiaoyuWen/MAGIC-Qwen2.5-14B-Instruct Text Generation • Updated Feb 4 • 4 • 1 XiaoyuWen/MAGIC-Qwen2.5-7B-Instruct Text Generation • 8B • Updated Feb 4 • 17 • • 2 XiaoyuWen/MAGIC-Llama3.1-8B-Instruct Text Generation • Updated Feb 4 • 8 XiaoyuWen/MAGIC Text Generation • Updated Feb 4 • 1