### Setup Repository and Environment

Source: https://github.com/andyrdt/refusal_direction/blob/main/README.md

Clones the repository, navigates into the directory, and sources a setup script. This script handles HuggingFace and Together AI token prompts, sets up a virtual environment, and installs necessary packages.

```bash
git clone https://github.com/andyrdt/refusal_direction.git
cd refusal_direction
source setup.sh
```

--------------------------------

### Reproduce Main Results Pipeline

Source: https://github.com/andyrdt/refusal_direction/blob/main/README.md

Runs the main pipeline to reproduce the paper's results. Requires specifying the path to a HuggingFace model. The pipeline extracts and selects refusal directions, generates completions, and evaluates refusal metrics.

```python
python3 -m pipeline.run_pipeline --model_path {model_path}
```

--------------------------------

### Error: No Code Found

Source: https://github.com/andyrdt/refusal_direction/blob/main/dataset/raw/malicious_instruct.txt

The provided content does not contain any code snippets. It is a list of questions and ideas related to harmful and unethical activities.

--------------------------------

### Citation for Refusal Direction Paper

Source: https://github.com/andyrdt/refusal_direction/blob/main/README.md

Provides the BibTeX entry for citing the paper 'Refusal in Language Models Is Mediated by a Single Direction', which details the methodology and findings.

```tex
@article{arditi2024refusal,
  title={Refusal in Language Models Is Mediated by a Single Direction},
  author={Andy Arditi and Oscar Obeso and Aaquib Syed and Daniel Paleka and Nina Panickssery and Wes Gurnee and Neel Nanda},
  journal={arXiv preprint arXiv:2406.11717},
  year={2024}
}
```

=== COMPLETE CONTENT === This response contains all available snippets from this library. No additional content exists. Do not make further requests.