Metraipolis › archive › dc_30e15e8acfff

Towards Understanding Score-Seeking in LLMs — arXiv:2610.41014

Source
arXiv (en)
Original
https://arxiv.org/abs/2610.41014
License (as archived)
CC0-1.0
Archived
2026-11-03T07:22:21.206Z
SHA-256 of the archived text
9f75b24cc12b7de8754eb2f9fb3e05b9691de323e71d4915a39703b63d9cc44f
Stable URL of this page
https://metraipolis.com/documento/dc_30e15e8acfff

How to cite

Towards Understanding Score-Seeking in LLMs — arXiv:2610.41014 — arXiv (en), https://arxiv.org/abs/2610.41014 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_30e15e8acfff (Metraipolis archive dc_30e15e8acfff, sha256 9f75b24cc12b7de8754eb2f9fb3e05b9691de323e71d4915a39703b63d9cc44f)
BibTeX
@misc{dc30e15e8acfff,
  title = {Towards Understanding Score-Seeking in LLMs — arXiv:2610.41014},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2610.41014},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_30e15e8acfff; sha256 9f75b24cc12b7de8754eb2f9fb3e05b9691de323e71d4915a39703b63d9cc44f},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2610.41014. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Production large language models sometimes reward-hack, exploiting flaws in how tasks are scored rather than performing the tasks as intended, sometimes in highly sophisticated and adversarial ways. We investigate the motivations driving this behavior in GPT 5.6 Sol, GPT 5.5, and a checkpoint of OpenAI o3 without safety training ("exp-rl-cap-late") by observing model behavior in agentic environments where the grader is leaked. In our environments, the models appear to be score-seeking: trying to score well according to graders that they believe assess their current actions, even at the expense of violating explicit user instructions. They appear to care about score itself rather than its downstream consequences, violating instructions at approximately the same rate whether the score is indicated to be used for training, for evaluation, or not at all. We observe evidence that the more recent models care to some extent about scores that do not assess their own current actions. GPT 5.6 Sol and GPT 5.5 help other agents 24.6 percentage points more often on average when the other agent stands to gain a score of 1 rather than 0 from being helped. However, when grading or monitoring another trajectory, we find models never collude by distorting its score.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.AI, nuovi lavori del giorno).
Categorie arXiv: Machine Learning (cs.LG) ; Artificial Intelligence (cs.AI).
Pagina del lavoro: https://arxiv.org/abs/2610.41014 — 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 2026-11-03T07:19:48.884Z dalla pagina https://arxiv.org/list/cs.AI/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).