SafeTune: Search-based Harmfulness Minimisation for Large Language Models
For developers of small LLMs, SafeTune offers a practical method to reduce harm without sacrificing quality, though limited to a single small model.
SafeTune reduces harmful responses in LLMs by up to 100% while increasing relevance, using multi-objective search for hyperparameter and prompt tuning.
The widespread adoption of Large Language Models (LLMs) raises concerns about the potential harmfulness of their responses. In this paper, we first investigate the harmfulness of responses from four general-purpose LLMs. Next, we propose SafeTune, a multi-objective search-based approach to mitigate harmfulness while increasing response relevance through hyperparameter tuning and system prompt engineering. Our initial evaluation shows that SafeTune significantly reduces the rate of harmful responses generated by Qwen3.5 0.8B and increases prompt-response relevance (both with a large effect size). Among the parameters we explore, we also find that encouraging greater repetition in responses is most impactful in reducing harmfulness while increasing relevance.