Metraipolis › archive › dc_7dfaea60356a

Distinguishing among reasoning levels of GPT 5.5 Pro via a grammar lens benchmark — arXiv:2611.00555

Source: arXiv (en)
Original: https://arxiv.org/abs/2611.00555
License (as archived): CC0-1.0
Archived: 2026-11-03T11:09:44.075Z
SHA-256 of the archived text: d2d7a5aebedc70a4453ea36c32b18d0f2295bd84941e6f826cf3ac1b1ad405d2
Stable URL of this page: https://metraipolis.com/documento/dc_7dfaea60356a

How to cite

Distinguishing among reasoning levels of GPT 5.5 Pro via a grammar lens benchmark — arXiv:2611.00555 — arXiv (en), https://arxiv.org/abs/2611.00555 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_7dfaea60356a (Metraipolis archive dc_7dfaea60356a, sha256 d2d7a5aebedc70a4453ea36c32b18d0f2295bd84941e6f826cf3ac1b1ad405d2)

BibTeX

@misc{dc7dfaea60356a,
  title = {Distinguishing among reasoning levels of GPT 5.5 Pro via a grammar lens benchmark — arXiv:2611.00555},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.00555},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_7dfaea60356a; sha256 d2d7a5aebedc70a4453ea36c32b18d0f2295bd84941e6f826cf3ac1b1ad405d2},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.00555. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Does the level of test-time processing effort affect a language model's computational capability? A formal grammar-lens benchmark method, derived from the theory of computation, generates novel data at specified levels, each successively more computationally powerful than the levels below. Using this data, we tested GPT-5.5, and GPT-5.5-Pro-Medium, -High, and -XHigh, showing incrementally increased acceptance of test data: 16, 40, 43, and 52\%, respectively, illustrating the benchmark's ability to discern fine-grained differences among models' added test-time reasoning levels. Additionally, as test-time compute effort increases, model responses diverge from human responses to the same data. This increasing disagreement may turn out to be broadly characteristic of more capable language models, introducing the potential of a general capability-versus-alignment conflict.

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/2611.00555 — 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 2026-11-03T10:59:52.259Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).