English

STAR: SocioTechnical Approach to Red Teaming Language Models

Artificial Intelligence 2024-10-24 v4 Computation and Language Computers and Society Human-Computer Interaction

Abstract

This research introduces STAR, a sociotechnical framework that improves on current best practices for red teaming safety of large language models. STAR makes two key contributions: it enhances steerability by generating parameterised instructions for human red teamers, leading to improved coverage of the risk surface. Parameterised instructions also provide more detailed insights into model failures at no increased cost. Second, STAR improves signal quality by matching demographics to assess harms for specific groups, resulting in more sensitive annotations. STAR further employs a novel step of arbitration to leverage diverse viewpoints and improve label reliability, treating disagreement not as noise but as a valuable contribution to signal quality.

Keywords

Cite

@article{arxiv.2406.11757,
  title  = {STAR: SocioTechnical Approach to Red Teaming Language Models},
  author = {Laura Weidinger and John Mellor and Bernat Guillen Pegueroles and Nahema Marchal and Ravin Kumar and Kristian Lum and Canfer Akbulut and Mark Diaz and Stevie Bergman and Mikel Rodriguez and Verena Rieser and William Isaac},
  journal= {arXiv preprint arXiv:2406.11757},
  year   = {2024}
}

Comments

8 pages, 5 figures, 5 pages appendix. * denotes equal contribution

R2 v1 2026-06-28T17:08:58.949Z