Researchers Open-Source LLM Jailbreak Defense Algorithm SafeDecoding

Researchers from the University of Washington, the Pennsylvania State University, and Allen Institute for AI have open-sourced SafeDecoding, a technique for protecting large language models (LLMs) against jailbreak attacks. SafeDecoding outperforms baseline jailbreak defenses without incurring significant computational overhead.

ai algorithm allen allen institute allen institute for ai anthony attacks computational deep learning defense generative-ai institute jailbreak language language models large language large language models llm llms ml & data engineering neural networks researchers state university university of washington washington