Human Language Shapes the Emerging Physical Reasoning in Large Language Models

Abstract

Whether large language models (LLMs) acquire adequate representations of physical reality by predicting the next token remains unresolved. Molecular conformations provide a quantitative test: we asked 28 LLMs to rank 810 geometries of 27 molecules from atomic coordinates alone. Frontier LLMs approach the accuracy of specialized force fields, but their ability is bound to well-defined chemical concepts. They rank conformers with more hydrogen bonds as more stable, inverting the true order when the hydrogen-bonded conformer is less stable. On pairs with equal hydrogen-bond counts, they are most accurate for molecules whose explanations invoke hydrogen bonding, whereas molecules governed by ring strain remain difficult. Thus, the physical understanding emerging in LLMs inherits the structure and blind spots of human scientific discourse, a machine analogue of linguistic relativity.

How Chemical Evidence Shapes Mechanistic Reasoning and Prediction by Large Language Models: Fine-Grained Inference in Nucleoside-Derived Supramolecular Gelation
Ying Wang, Xi Yu
ChemRxiv3 Nov 2026PreprintLicence not statedGeneral chemistryChemRxiv
Surfactant-Toolkit and SurfBench-200: an open computation toolkit and a benchmark of tool-augmented language-model reasoning in surfactant science
Prashant D. Kotian
ChemRxiv2 Nov 2026PreprintLicence not statedGeneral chemistryChemRxiv
Hierarchical Multi-agent Large Language Model Reasoning for Autonomous Heterogeneous Catalyst Discovery
Samuel Rothfarb, Megan Christina Davis, Ivana Matanović +3
npj Computational Materials25 May 2026Peer-reviewedCC BY-NC-NDFull text8 citedMaterials & polymersOpenAlex
Separating perception from reasoning in vision-language models: a model-free render ceiling for crystal structures
Can Polat, Mustafa Kurban, Erchin Serpedin +1
arXiv1 Sep 2026PreprintCC BY 4.0Full textPhysical chemistry Condensed matterarXiv
A Zero-Shot Single-Point Chemical Language Model Mimics Medicinal Chemist Reasoning for Molecular Optimization and STING Inhibitor Development in Alzheimer’s Disease
Peng Gao, Ying Qin, Zhilian Dai +10
ChemRxiv8 Sep 2026PreprintCC BYGeneral chemistryChemRxiv

Cite

BibTeX RIS

BibTeX

@misc{semakin2026human,
  title = {{Human Language Shapes the Emerging Physical Reasoning in Large Language Models}},
  author = {Gregorii A. Semakin and Timofey V. Losev and Ilya V. Prolomov and Stepan N. Ostarkov and Igor V. Alabugin and Michael G. Medvedev},
  howpublished = {arXiv},
  year = {2026},
  month = oct,
  doi = {10.48550/arxiv.2609.20666},
  eprint = {2609.20666},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2609.20666},
}

RIS

TY  - UNPB
TI  - Human Language Shapes the Emerging Physical Reasoning in Large Language Models
AU  - Gregorii A. Semakin
AU  - Timofey V. Losev
AU  - Ilya V. Prolomov
AU  - Stepan N. Ostarkov
AU  - Igor V. Alabugin
AU  - Michael G. Medvedev
PY  - 2026
DA  - 2026/10/27
PB  - arXiv
DO  - 10.48550/arxiv.2609.20666
UR  - https://arxiv.org/abs/2609.20666
AB  - Whether large language models (LLMs) acquire adequate representations of physical reality by predicting the next token remains unresolved. Molecular conformations provide a quantitative test: we asked 28 LLMs to rank 810 geometries of 27 molecules from atomic coordinates alone. Frontier LLMs approach the accuracy of specialized force fields, but their ability is bound to well-defined chemical concepts. They rank conformers with more hydrogen bonds as more stable, inverting the true order when the hydrogen-bonded conformer is less stable. On pairs with equal hydrogen-bond counts, they are most accurate for molecules whose explanations invoke hydrogen bonding, whereas molecules governed by ring strain remain difficult. Thus, the physical understanding emerging in LLMs inherits the structure and blind spots of human scientific discourse, a machine analogue of linguistic relativity.
ER  -

Abstract

Related

Cite