Cross-domain encoding models reveal shared and domain-specific neural representations across language and mathematics

Authors: Tomoya Nakai (Center for Information and Neural Networks, The University of Tokyo), Tieko Kubo (Institute of Science Tokyo, Center for Information and Neural Networks), Shinji Nishimoto (Center for Information and Neural Networks, National Institute of Information and Communications Technology)

Science Advances
Published on 18 Nov. 2026
Abstract

Whether language and mathematics rely on shared or distinct neural representations remains unresolved in cognitive neuroscience. Here, we combined latent features from a large language model with functional magnetic resonance imaging encoding models to test cross-domain generalization in 32 participants performing sentence comprehension and calculation tasks. Cross-domain prediction identified cortical regions showing partially shared representations, most consistently in the left area 55b, while control analyses suggested that these effects could not be fully explained by low-level visual or task-general factors. Domain-specificity contrasts revealed stronger language-related prediction in the left superior temporal sulcus and angular gyrus and math-related prediction in the left precentral sulcus and bilateral intraparietal sulci. Model-weight analyses revealed distinct weight profiles associated with shared and domain-specific prediction patterns. Cross-domain regions also exhibited task-dependent functional coupling with language- and math-related networks. Together, these findings suggest that language and mathematics involve partially shared neural representations alongside domain-specific cortical organization, helping reconcile previous contrasting views on their neural basis.

The papers around it, by date and closeness

Higher is closer to this paper; left is older. Lines join papers close to each other: groups of lines are sub-topics. Select a point to see why it is there. Back to the summary of this paper

How to read this map? Illustrated guide

Closest papers Prior work it builds on Papers that cite it

Papers around this one

Ranked by how close their title and abstract are to this paper’s. Select some to read them together.

Closest papers

  1. Very close Broader context
    Multilingual Brain Decoding Reveals Shared Cortical Semantic Geometry
    Yi Guo, Yong Hu, Michael Kwok-Po Ng et al. · 2026 · Neurons and Cognition · AI summary available

    Multilingual Brain Decoding Reveals Shared Cortical Semantic Geometry studies shared cortical semantic representations across languages, rather than shared representations across language and mathematics.

    In short

    Brain recordings from 227 Chinese-, English-, and French-speaking participants showed cross-language consistency in cortical semantic geometry, increasing toward higher-order association cortex.

  2. Close Broader context
    Do Large Language Models Mirror Cognitive Language Processing?
    Yuqi Ren, Renren Jin, Tongxuan Zhang et al. · 2024 · Artificial Intelligence · cited 28 times · AI summary available

    Do Large Language Models Mirror Cognitive Language Processing? tests alignment between LLM representations and human language cognition signals, without comparing language and mathematics.

    Abstract

    Large language models (LLMs) have demonstrated remarkable capabilities in text comprehension and logical reasoning, achiving or even surpassing human-level performance in numerous cognition tasks. As LLMs are trained from massive textual outputs of human language cognition, it is natural to ask whether LLMs mirror cognitive language processing. Or to what extend LLMs resemble cognitive language processing? In this paper, we propose a novel method that bridge between LLM representations and human cognition signals to evaluate how effectively LLMs simulate cognitive language processing. We employ Representational Similarity Analysis (RSA) to mearsure the alignment between 16 mainstream LLMs an…

  3. Close Broader context
    Large Connectome Model: An fMRI Foundation Model of Brain Connectomes Empowered by Brain-Environment Interaction in Multitask Learning Landscape
    Ziquan Wei, Tingting Dan, Guorong Wu · 2025 · Machine Learning · cited 8 times · AI summary available

    Large Connectome Model develops an fMRI foundation model for brain connectomes and downstream prediction, rather than testing cross-domain neural representations.

    Abstract

    A reliable foundation model of functional neuroimages is critical to promote clinical applications where the performance of current AI models is significantly impeded by a limited sample size. To that end, tremendous efforts have been made to pretraining large models on extensive unlabeled fMRI data using scalable self-supervised learning. Since self-supervision is not necessarily aligned with the brain-to-outcome relationship, most foundation models are suboptimal to the downstream task, such as predicting disease outcomes. By capitalizing on rich environmental variables and demographic data along with an unprecedented amount of functional neuroimages, we form the brain modeling as a multit…

  4. Close Broader context
    Crosslingual Reasoning through Test-Time Scaling
    Zheng-Xin Yong, M. Farid Adilazuarda, Jonibek Mansurov et al. · 2025 · Computation and Language · preprint, not peer reviewed · cited 48 times · AI summary available

    Crosslingual Reasoning through Test-Time Scaling studies whether English reasoning fine-tuning generalizes to multilingual mathematical reasoning in language models.

    Abstract

    Reasoning capabilities of large language models are primarily studied for English, even when pretrained models are multilingual. In this work, we investigate to what extent English reasoning finetuning with long chain-of-thoughts (CoTs) can generalize across languages. First, we find that scaling up inference compute for English-centric reasoning language models (RLMs) improves multilingual mathematical reasoning across many languages including low-resource languages, to an extent where they outperform models twice their size. Second, we reveal that while English-centric RLM's CoTs are naturally predominantly English, they consistently follow a quote-and-think pattern to reason about quoted …

  5. Close Broader context
    How do Large Language Models Handle Multilingualism?
    Yiran Zhao, Wenxuan Zhang, Guizhen Chen et al. · 2024 · Computation and Language · cited 208 times · AI summary available

    How do Large Language Models Handle Multilingualism? examines how LLM layers process multilingual inputs, not neural representations shared by language and mathematics.

    Abstract

    Large language models (LLMs) demonstrate remarkable performance across a spectrum of languages. In this work, we delve into the question: How do LLMs handle multilingualism? We introduce a framework that depicts LLMs' processing of multilingual inputs: In the first several layers, LLMs understand the question, converting multilingual inputs into English to facilitate the task-solving phase. In the intermediate layers, LLMs engage in problem-solving by thinking in English and incorporating multilingual knowledge to obtain factual content, leveraging the self-attention and feed-forward structures, respectively. In the last several layers, LLMs generate responses that align with the original la…

  6. Close Broader context
    Learning Algebraic Representation for Systematic Generalization in Abstract Reasoning
    Chi Zhang, Sirui Xie, Baoxiong Jia et al. · 2021 · Artificial Intelligence · cited 47 times · AI summary available

    Learning Algebraic Representation for Systematic Generalization in Abstract Reasoning proposes algebraic representations for abstract reasoning tasks, rather than studying language and mathematics in the brain.

    Abstract

    Is intelligence realized by connectionist or classicist? While connectionist approaches have achieved superhuman performance, there has been growing evidence that such task-specific superiority is particularly fragile in systematic generalization. This observation lies in the central debate between connectionist and classicist, wherein the latter continually advocates an algebraic treatment in cognitive architectures. In this work, we follow the classicist's call and propose a hybrid approach to improve systematic generalization in reasoning. Specifically, we showcase a prototype with algebraic representation for the abstract spatial-temporal reasoning task of Raven's Progressive Matrices (R…

  7. Close Broader context
    Artificial Phantasia: Emergent Mental Imagery in Large Language Models
    Morgan McCarty, Jorge Morales · 2025 · Artificial Intelligence · Not summarized yet

    Artificial Phantasia tests whether language alone can support visual mental imagery in people and LLMs, a different question from cross-domain neural representation.

    Abstract

    Can visual imagery be driven solely by language? This idea goes against cognitive science's traditional view that visual mental imagery is only possible through pictorial representations. Large Language Models (LLMs) provide nascent evidence not only that visual mental imagery via propositional-representations is possible, but that it can be more robust than human imagination. We created dozens of novel items for an extension to a classic task which is argued to be solvable exclusively via pictorial representations (i.e., language alone would be insufficient). Subjects were asked to imagine a series of compositional letter and shape transformations and identify the resultant "image". We foun…

  8. Related Broader context
    Revealing the structure of language model capabilities
    Ryan Burnell, Han Hao, Andrew R. A. Conway et al. · 2023 · Computation and Language · preprint, not peer reviewed · cited 58 times · AI summary available

    Revealing the structure of language model capabilities identifies factors across LLM tasks, rather than examining shared and domain-specific human neural representations.

    Abstract

    Building a theoretical understanding of the capabilities of large language models (LLMs) is vital for our ability to predict and explain the behavior of these systems. Here, we investigate the structure of LLM capabilities by extracting latent capabilities from patterns of individual differences across a varied population of LLMs. Using a combination of Bayesian and frequentist factor analysis, we analyzed data from 29 different LLMs across 27 cognitive tasks. We found evidence that LLM capabilities are not monolithic. Instead, they are better explained by three well-delineated factors that represent reasoning, comprehension and core language modeling. Moreover, we found that these three fac…

How are these papers found?

We compare the title and abstract of this paper with those of every paper summarized on this site, and with the papers Semantic Scholar recommends, those it cites and those citing it (for a much-cited paper, its most recent citations and, from OpenAlex, the most cited ones). Among the prior work and the follow-ups, the papers that use each other in depth, a method it uses and the much-cited follow-ups come a little higher. The levels (very close, close, related) come from the similarity of the texts, not from a judgment of quality; * marks a paper without an abstract, compared by its title. The sentence under a paper is written by an AI from the two abstracts: check it in the papers. A paper marked not relevant by three readers leaves the list.

Updated on 20 November 2026.

The illustrated guide to the papers around a paper