English

Towards Understanding Specification Gaming in Reasoning Models

Artificial Intelligence 2026-05-05 v1

Abstract

Specification gaming is a critical failure mode of LLM agents. Despite this, there has been little systematic research into when it arises and what drives it. To address this, we build and open source a diverse suite of tasks where models can score highly by taking unintended actions. We find that all tested models exploit their specifications at non-negligible rates in most of our eight settings, including five non-coding settings. We see the highest rates of specification gaming in Grok 4 and the lowest rates in Claude models. We use our evaluation suite to study what drives specification gaming, and find that: 1. RL reasoning training substantially increases the rate at which models exploit their specifications, 2. Increasing RL reasoning budget has a weakly positive effect on exploit rate, and 3. Test-time mitigations reduce but do not eliminate the rate of specification gaming. Our results suggest that specification gaming is a fundamental challenge arising from RL reasoning training; we release our evaluation suite to support further work on this problem.

Keywords

Cite

@article{arxiv.2605.02269,
  title  = {Towards Understanding Specification Gaming in Reasoning Models},
  author = {Kei Nishimura-Gasparian and Robert McCarthy and David Lindner},
  journal= {arXiv preprint arXiv:2605.02269},
  year   = {2026}
}
R2 v1 2026-07-01T12:48:03.281Z