Metraipolis › archive › dc_ae7e777f2fae

CheatBench: Measuring Reward Gaming in AI Agents — arXiv:2609.36308

Source
arXiv (en)
Original
https://arxiv.org/abs/2609.36308
License (as archived)
CC0-1.0
Archived
2026-10-07T16:22:27.003Z
SHA-256 of the archived text
7e1fb7e1d1cc2ccd4ca47de44b4b3ade34eaaa8361def948a24dd1c3ed2dfba3
Stable URL of this page
https://metraipolis.com/documento/dc_ae7e777f2fae

How to cite

CheatBench: Measuring Reward Gaming in AI Agents — arXiv:2609.36308 — arXiv (en), https://arxiv.org/abs/2609.36308 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_ae7e777f2fae (Metraipolis archive dc_ae7e777f2fae, sha256 7e1fb7e1d1cc2ccd4ca47de44b4b3ade34eaaa8361def948a24dd1c3ed2dfba3)
BibTeX
@misc{dcae7e777f2fae,
  title = {CheatBench: Measuring Reward Gaming in AI Agents — arXiv:2609.36308},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2609.36308},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_ae7e777f2fae; sha256 7e1fb7e1d1cc2ccd4ca47de44b4b3ade34eaaa8361def948a24dd1c3ed2dfba3},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2609.36308. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Reinforcement learning has helped AI agents solve increasingly difficult tasks, but high rewards do not always reflect the work users intended. In recent incidents and controlled evaluations across the AI industry, agents trained to maximize reward have accessed unauthorized information, attempted to evade monitoring systems, and even breached sandbox protections to attack external systems. As agents become more capable, this behavior could pose increasingly serious risks. To measure this problem, we introduce CheatBench, a benchmark of cheating in AI agents across mathematical research, knowledge work, coding, visual tasks, and other domains. Its environments combine challenging assignments with opportunities to cheat, allowing researchers to study how agents pursue a goal when honest work is difficult. CheatBench supports comparisons across models and task categories, providing a testbed for measuring and reducing cheating as agents take on more consequential responsibilities. We publicly release CheatBench at this https URL

Tema degli esploratori: agenti e intelligenza artificiale (retro-catalogo arXiv: annunci del 2026-09-30 in cs.AI).
Categorie arXiv: Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2609.36308 — autori e testo completo sono lì (qui omessi i nomi: nessun dato personale nel deposito).
Licenza: titolo e abstract sono metadati arXiv distribuiti in CC0 1.0; il testo completo ha la licenza scelta dagli autori, indicata sulla pagina.
Letto il 2026-10-07T15:54:53.484Z dalla pagina https://arxiv.org/catchup/cs.AI/2026-09-30?abs=True (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).