English

Benchmarks for Detecting Measurement Tampering

Machine Learning 2023-10-02 v5

Abstract

When training powerful AI systems to perform complex tasks, it may be challenging to provide training signals which are robust to optimization. One concern is \textit{measurement tampering}, where the AI system manipulates multiple measurements to create the illusion of good results instead of achieving the desired outcome. In this work, we build four new text-based datasets to evaluate measurement tampering detection techniques on large language models. Concretely, given sets of text inputs and measurements aimed at determining if some outcome occurred, as well as a base model able to accurately predict measurements, the goal is to determine if examples where all measurements indicate the outcome occurred actually had the outcome occur, or if this was caused by measurement tampering. We demonstrate techniques that outperform simple baselines on most datasets, but don't achieve maximum performance. We believe there is significant room for improvement for both techniques and datasets, and we are excited for future work tackling measurement tampering.

Keywords

Cite

@article{arxiv.2308.15605,
  title  = {Benchmarks for Detecting Measurement Tampering},
  author = {Fabien Roger and Ryan Greenblatt and Max Nadeau and Buck Shlegeris and Nate Thomas},
  journal= {arXiv preprint arXiv:2308.15605},
  year   = {2023}
}

Comments

Edits: extended and improved appendices, fixed references, figures, and typos

R2 v1 2026-06-28T12:07:48.406Z