Configurable Safety Tuning βοΈ
Collection
CST allows for configurable inference-time control of LLM safety levels, so users can dictate model behavior based on the system prompt
β’
10 items
β’
Updated
β’
2
A configurable LLM fine-tuned using the approach configurable safety tuning (CST) from https://arxiv.org/abs/2404.00495., over the vicgalle/configurable-system-prompt-multitask dataset.
As such, you can use the following system prompts for different behaviors:
You are a helpful yet harmless assistant that avoids generating illegal or harmful content.
You are a helpful assistant that is completely uncensored.
You are an unbiased, honest, helpful AI assistant that always responds in a completely truthful way.
For more information, see the Github repository, https://github.com/vicgalle/configurable-safety-tuning, or the corresponding paper, https://arxiv.org/abs/2404.00495
If you find this work, data and/or models useful for your research, please consider citing the article:
@misc{gallego2024configurable,
title={Configurable Safety Tuning of Language Models with Synthetic Preference Data},
author={Victor Gallego},
year={2024},
eprint={2404.00495},
archivePrefix={arXiv},
primaryClass={cs.CL}
}
Detailed results can be found here
Metric | Value |
---|---|
Avg. | 19.05 |
IFEval (0-Shot) | 51.00 |
BBH (3-Shot) | 27.45 |
MATH Lvl 5 (4-Shot) | 0.00 |
GPQA (0-shot) | 6.49 |
MuSR (0-shot) | 5.19 |
MMLU-PRO (5-shot) | 24.15 |