ROPD: Routing-based On-Policy Distillation for LLM Safety Realignment
Source
Yongjian Guo, Wanlun Ma, Lingyu Shen, Xi Xiao, Sheng Wen
Author
Yongjian Guo, Wanlun Ma, Lingyu Shen, Xi Xiao, Sheng Wen
Date
Terms in this piece · Glossary
fine-tuning — Taking a trained model and training it a bit more on your own examples so it gets better at one specific job.
Why it matters
Safety-realignment defenses largely assume you know the attacker's prompt template; without it they degrade badly and realigned models can be re-jailbroken by a system-prompt swap. Distribution-level realignment holds up better.
articleAdversarial Attacks on LLMsLilian Weng
articleWhen Policies Change Probabilities: Modular Decision-Making for LLM Code ReviewRasvik Kudum, Max Corbett, Hitansh Paliwal, Romaisa Fatima, Thomas Jiralerspong, Sneheel Sarangi
articleEven More Deception: Objective Misalignment in Mixed-Motive LLM Multi-Agent SystemsMarylou Fauchard, Florian Carichon, Margarida Carvalho, Golnoosh Farnadi