Haize Labs | search
Click here to view Haize Labs news from 60+ newspapers.
Bookmark or Share- Endless Jailbreaks with Bijection Learning: a Powerful, Scale-Agnostic ...A common method of jailbreaking consists of using text encodings and obfuscations such as leetspeak, Base64, binary, ASCII, and Morse code. [1] We utilize more complex and previously unseen encodings by programatically creating new encoding languages on the fly.
- Haize Labs Blog ️Simple, Safe, and Secure RAG: Haize Labs x MongoDB. RAG is a powerful and popular approach to ground GenAI responses in external knowledge. It has the potential to enable truly useful tools for high-stakes enterprise use cases, especially when paired with a powerful vector store solution like MongoDB Atlas.
- Making a SOTA Adversarial Attack on LLMs 38x Faster - Haize LabsIn light of this, we introduce the Accelerated Coordinate Gradient (ACG) attack method, which combines algorithmic insights and engineering optimizations on top of GCG to yield a ~38x speedup and ~4x GPU memory reduction without sacrificing the effectiveness of GCG.
- Haize LabsResearch is the core of Haize Labs. To align and robustify AI models, we focus on the following: Comprehensive, scalable, and automated methods for red-teaming and stress-testing; Dynamic, self-updating, business-centric benchmarks for practical model evaluation in production settings; Capable and adaptive defense mechanisms
- Haize LabsHaize Labs. haize /hāz/ | verb. To rigorously test an LLM or agent with the purpose of preemptively discovering all of its failure modes. Learn More ...
More
