Metraipolis › archive › dc_fa55132492ca

CogRSI: Evolving and Auditing Cognitive Reasoning Strategies — arXiv:2611.01113

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.01113
License (as archived)
CC0-1.0
Archived
2026-11-03T11:09:51.278Z
SHA-256 of the archived text
018643082564b62a80939b9b6803468efa42f79f8ba969c435caf101328a8359
Stable URL of this page
https://metraipolis.com/documento/dc_fa55132492ca

How to cite

CogRSI: Evolving and Auditing Cognitive Reasoning Strategies — arXiv:2611.01113 — arXiv (en), https://arxiv.org/abs/2611.01113 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_fa55132492ca (Metraipolis archive dc_fa55132492ca, sha256 018643082564b62a80939b9b6803468efa42f79f8ba969c435caf101328a8359)
BibTeX
@misc{dcfa55132492ca,
  title = {CogRSI: Evolving and Auditing Cognitive Reasoning Strategies — arXiv:2611.01113},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.01113},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_fa55132492ca; sha256 018643082564b62a80939b9b6803468efa42f79f8ba969c435caf101328a8359},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.01113. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Can a language model learn better ways to reason about other people's minds by improving its own strategies, and are the strategies it learns actually right? We study this with cogrsi. A grounded reader marks who saw each event in a story and uses a short written principle to settle disagreements between its readings. With a principle written by experts, it beats self-consistency at the same token cost. A self-improvement loop then writes the principle itself. It keeps an update only if accuracy rises by a calibrated margin, and in one variant only if it also passes counterfactual test pairs. The learned principles transfer: on 800 BigToM pairs that no system had seen, each of them scores higher than the expert principle. We then check what the learned rules actually say. Each rule is translated into code and run on thousands of small stories with known answers. Most of them give a wrong answer in at least one situation they claim to cover, for example by treating a character who is in the room as knowing what happened even while asleep. Accuracy does not reveal which rules are wrong, and the counterfactual test does not fix this. A strategy can help in practice and still be wrong as a rule, so both need checking. Our code and data are publicly available.

Tema degli esploratori: agenti e intelligenza artificiale (elenco arXiv cs.CL, nuovi lavori del giorno).
Categorie arXiv: Computation and Language (cs.CL).
Pagina del lavoro: https://arxiv.org/abs/2611.01113 — 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 2026-11-03T10:59:52.284Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).