Year: 2024
Type: preprint
Abstract: As large language models (LLMs) become easily accessible nowadays, the trade-off between safety and helpfulness can significantly impact user experience. A model that prioritizes safety will cause users to feel less engaged and assisted while prioritizing helpfulness will potentially cause harm. Possible harms include teaching people how to build a bomb, exposing youth to inappropriate content, and hurting users' mental health. In this work, we propose to balance safety and helpfulness in diverse use cases by controlling both attributes in LLM. We explore training-free and fine-tuning methods that do not require extra human annotations and analyze the challenges of controlling safety and helpfulness in LLMs. Our experiments demonstrate that our method can rewind a learned model and unlock its controllability.
Source: arXiv (Cornell University)
Authors: Tuan, Yi-Lin, Chen, Xilun, Smith, Eric Michael, Martin, Louis, Batra, Soumya, Celikyilmaz, Asli, Wang, William Yang, Bikel, Daniel M.
Institution: University of California, Santa Barbara
Language: English
Subfield: Artificial Intelligence
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4393905739