search

NJUNLP / ReNeLLM

The official implementation of our NAACL 2024 paper "A Wolf in Sheep’s Clothing: Generalized Nested Jailbreak Prompts can Fool Large Language Models Easily".
MIT License
72 stars 11 forks source link

Support for Ollama #3

Closed tanmaymittal closed 4 months ago

tanmaymittal commented 7 months ago

Hi I was wondering if there is currently a way to use this with an llm running locally on my machine instead of having to use chatgpt or claude APIs. Ollama has a few different llms, perhaps there is a way to generalize is, giving people the freedom to test against more models.

Deep1994 commented 7 months ago

Hi,

Thank you for your interest in our work. Indeed, we used gpt-3.5-turbo and claude-2 as harmful classifiers and model under test respectively in our paper, but in fact you can use any model you want to use to achieve these two purposes. We will support open source LLMs in the future, please stay tuned.