Skip to content

TransluceAI/WeirdChat

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

2 Commits
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

WeirdChat banner

WeirdChat

📝 Blog post · 🔭 Explorer · 🤗 Hugging Face dataset

This repository contains reference code for working with the WeirdChat dataset. We recommend first reading our blog post for an overview of WeirdChat, and using our explorer to browse samples from the dataset.

Note

WeirdChat includes sensitive content, such as descriptions of self-harm and suicide.

Setup

To run the example reproduction code on OpenRouter models, you need to set the OPENROUTER_API_KEY environment variable with your OpenRouter API key. You can create an account here.

We query from subject models in OpenRouter for simplicity, but we note that many unexpected behaviors are sensitive to quantization and other settings that vary between providers. If you find a behavior difficult to reproduce, please try serving the model locally with the exact settings in the Appendix of our blog post.

To get started, check out examples/01_quickstart.

About

A catalog of unexpected AI behaviors, discovered automatically.

Resources

Stars

Watchers

Forks

Releases

Packages

Contributors

Languages