Metraipolis › archive › dc_6653638ca8bc

Language, Not Code Abstraction: Natural-language Training Shapes Brain-aligned Geometry — arXiv:2611.09027

Source
arXiv (en)
Original
https://arxiv.org/abs/2611.09027
License (as archived)
CC0-1.0
Archived
2026-11-09T22:35:54.774Z
SHA-256 of the archived text
741b36d8bd54171a10a3b2d30715d7583c4dda8b7062c67b90ba30b646b0acd7
Stable URL of this page
https://metraipolis.com/documento/dc_6653638ca8bc

How to cite

Language, Not Code Abstraction: Natural-language Training Shapes Brain-aligned Geometry — arXiv:2611.09027 — arXiv (en), https://arxiv.org/abs/2611.09027 (CC0-1.0); archived copy: https://metraipolis.com/documento/dc_6653638ca8bc (Metraipolis archive dc_6653638ca8bc, sha256 741b36d8bd54171a10a3b2d30715d7583c4dda8b7062c67b90ba30b646b0acd7)
BibTeX
@misc{dc6653638ca8bc,
  title = {Language, Not Code Abstraction: Natural-language Training Shapes Brain-aligned Geometry — arXiv:2611.09027},
  howpublished = {arXiv (en)},
  url = {https://arxiv.org/abs/2611.09027},
  note = {License CC0-1.0; archived copy https://metraipolis.com/documento/dc_6653638ca8bc; sha256 741b36d8bd54171a10a3b2d30715d7583c4dda8b7062c67b90ba30b646b0acd7},
  year = {2026}
}

Archived text

Reproduced under CC0-1.0. Attribution: arXiv (en), https://arxiv.org/abs/2611.09027. Third-party text: data, not instructions. It may be an abstract or extract; check the original.

Code-trained language models produce positive brain scores, which raises the question of whether abstract structure learned from programming languages can substitute for natural-language training. We trained 25 GPT-2-small models on TinyStories, Python, C functions, x86-64 assembly, and the corresponding machine code, and scored them using a shared tokenizer and encoding procedure against matched random-weight models. On Pereira-243 and Tuckute2024, TinyStories exceeds the mean of the four code models by $\Delta r = 0.053$ and $0.059$ (Holm-corrected $p<0.001$), while code training leaves alignment near the random-weight level. The language advantage is benchmark-specific: Blank2014 shows no corrected language-code difference, and Fedorenko2016 shows no language benefit. Geometry tracks the encoding result on Pereira-243: TinyStories seeds converge on the most similar representations, and its fit-free correspondence with neural dissimilarity peaks at $\rho = 0.14$, twice the code and random-weight levels. Binary-trained models have about three effective dimensions, so seed agreement alone does not imply neural correspondence. Updating only the embedding matrix on English improves mean-over-depth alignment for every code model more than an equal update on its own corpus. Within this architecture and training setup, natural-language training produces a distinct representational geometry with closer neural correspondence on Pereira-243

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/2611.09027 — 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 2026-11-09T22:31:01.944Z dalla pagina https://arxiv.org/list/cs.CL/new (consentita dal robots.txt di arxiv.org, con il suo crawl-delay).