dreadnode
@dreadnode
AI Red Teaming | Research. Tooling. Evals. Cyber ranges.
ID:173341654
https://dreadnode.io 01-08-2010 04:03:35
21 Tweet
818 Takipçi
23 Takip Edilen
I took an early stab at PGD for LLMs based on arxiv.org/abs/2402.09154 (Simon Geisler). Neat technique to relax the one-hot for gradient updates + projection. Also got to spend some time with litgpt.
github.com/dreadnode/rese…
Experimental and messy, but enjoy.
We've been on the road this last month - please enjoy the slides we can share, including a small workshop we gave Apres Cyber Slopes Summit
github.com/dreadnode/conf…
Was an awesome training environment. Props to SpecterOps for the work they’ve put into their RTO course.
So. Much. Content.
Would highly recommend to all aspiring red teamers.
Minimal implementation of the Tree of Attacks (TAP) LLM jailbreaking from Robust Intelligence:
github.com/dreadnode/parl…
- Cleaned + updated the prompts
- Uses OpenAI, Mistral, and TogetherAI APIs
- Refactored the leaf branching strategy