Abstract

Large language models (LLMs) routinely face requests that should be refused, creating a trade-off between helpfulness and harm prevention. However, refusals themselves can be helpful. In high-risk interactions involving crisis, coercion, or escalating intent, blunt non-compliance may prevent direct harm while still failing to support the needs of the person behind the request. We present PsychoSafe, a psychologically-informed refusal framework that reframes refusal as structured supportive communication grounded in evidence-based intervention strategies. To develop PsychoSafe, we construct a corpus of 8019 prompt-response pairs spanning five psychologically salient risk domains and apply prompting and parameter-efficient fine-tuning to Qwen 3.5 27B. On a balanced validation set of 500 prompts, evaluated with an LLM judge and validated through human ratings, PsychoSafe prompting improves overall refusal quality by 28.1% over a generic baseline, with particularly strong gains in external resource referral (+46.8%) and psychological grounding (+34.8%), while preserving downstream performance on non-refusal tasks. Fine-tuning achieves near-perfect refusal and resource-referral rates but reduces response relevance. Additional evaluations on SORRY-Bench and XSTest show strong in-domain robustness but limited out-of-domain generalization, suggesting that future work should diversify fine-tuning data to help models apply interventions selectively rather than schematically.

Citation

Gianluca Barmina, Federico Torrielli, Sven Harms, Jacob Nielsen, Felix Mächtle, Stine Lyngsø Beltoft, Peter Schneider-Kamp, Thomas Eisenbarth, Lukas Galke Poech, and Anne Lauscher, “PsychoSafe: Eliciting Psychologically-Informed Refusals in Large Language Models,” arXiv preprint arXiv:2606.09697, 2026. DOI: 10.48550/arXiv.2606.09697

@misc{barmina2026psychosafe,
 title        = {PsychoSafe: Eliciting Psychologically-Informed Refusals in Large Language Models},
 author       = {Gianluca Barmina and Federico Torrielli and Sven Harms and Jacob Nielsen and Felix M{\"a}chtle and Stine Lyngs{\o} Beltoft and Peter Schneider-Kamp and Thomas Eisenbarth and Lukas Galke Poech and Anne Lauscher},
 year         = 2026,
 eprint       = {2606.09697},
 archivePrefix = {arXiv},
 primaryClass = {cs.CL},
 doi          = {10.48550/arXiv.2606.09697},
 url          = {https://arxiv.org/abs/2606.09697}
}