Pith. sign in

REVIEW 3 major objections 5 minor 36 references

What Do Large Language Models Know? Tacit Knowledge as a Potential Causal-Explanatory Structure

T0 review · 3 major / 5 minor · reviewed 2026-08-16 · deepseek-v4-flash

Challenge this review Re-run · record.json Download PDF Read on arXiv ↗ Review JSON

Pith's one-line read This paper argues that at least some large language models can acquire tacit knowledge in Martin Davies's sense, because word embeddings supply the needed shared structure and ROME-style edits reveal causal common factors.

desk verdict A serious, clearly-written philosophy paper that applies Davies's tacit knowledge to LLMs via a weakened syntactic-structure constraint; the central bridge is asserted more than shown, but the paper deserves peer review. read the letter →

arxiv 2504.12187 v1 pith:NIABZENX submitted 2025-04-16 cs.CL cs.AIcs.LG

Céline Budding This is my paper · ORCID
classification cs.CLcs.AIcs.LG
keywords tacitknowledgelargelanguagemodelsDavies'sconstraintswordembeddingscausalcommonfactorsmodeleditingROMEsemantic
verification ladder T0 review T1 audit T2 compute T3 formal

The pith

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

The reading

This paper argues that at least some large language models can be said to know things in a real, if implicit, sense: they can acquire tacit knowledge as Martin Davies defined it, meaning rules that are not explicitly represented but that exist as causally relevant internal structures guiding behavior. This matters because it offers a middle path between saying LLMs merely conform to rules and saying they hold explicit knowledge, and it gives a principled target for intervention: edit the internal structure, and the whole family of related predictions changes. Davies himself concluded that connectionist networks fail the syntactic-structure constraint, but the paper contends that transformer LLMs meet a weakened version: the embedding layer maps semantically similar inputs to nearby vectors, which plays the functional role of Davies's shared syntactic property. The paper then reads the generalization and specificity results of ROME model-editing experiments as preliminary evidence that MLP key-value pairs act as Davies-style causal common factors, so that tacit knowledge is a workable conceptual tool for explaining and changing LLM behavior.

What carries the argument

The paper's load-bearing machinery is Davies's three-constraint test for tacit knowledge—semantic description, causal systematicity, and syntactic structure—together with two transformer components that let the test run on modern LLMs. The embedding layer is the key adaptor: by placing similar words and contexts near each other in a high-dimensional vector space, it performs the function Davies assigned to a shared syntactic property, so semantically similar inputs can engage the same causal processing without sharing an identical activation pattern. The candidate causal common factor is the MLP key-value pair, a feed-forward module in an intermediate layer that stores factual associations; when edited with the rank-one model editing (ROME) procedure, it updates predictions across paraphrases, which is the intervention signature expected of one causal structure mediating a whole family of related transitions. Generalization and specificity metrics from such edit experiments are the operational test for whether a candidate structure is a genuine causal common factor or merely a memorized one-to-one mapping.

What would settle it

Run a controlled edit experiment on a current LLM: select one hundred pairs of paraphrases whose embeddings are close, apply a rank-one edit to one member of each pair, and measure whether the other member's output changes; if close-embedding paraphrases often fail to co-update while the edited prompt itself changes, the claimed link between embedding proximity and a shared causal common factor is falsified.

Watch

Extended reading notes

Core claim

Stated on its own terms, the paper's claim is that Davies's account of tacit knowledge can be transferred to transformer LLMs with one modification, and that at least one current model appears to satisfy it. Semantic description holds because LLM inputs and outputs are word sequences with conventional meaning. Causal systematicity holds provisionally because the MLP key-value pairs localized by causal tracing in GPT-J behave the way Davies-style causal common factors should: a rank-one edit to one pair updates the prediction not only for the edited prompt but for semantically related paraphrases, while leaving unrelated prompts mostly unchanged, which is the generalization-and-specificity signature of a shared causal mechanism. The syntactic-structure constraint, which Davies thought distributed networks could not meet because context alters activation patterns, is weakened so that proximity in the embedding space replaces the requirement of an identical shared property; the embedding layer acts as a semantic categorizer that routes similar inputs through similar processing. The conclusion is that LLMs can meet all three modified constraints and that the GPT-J results give preliminary evidence that some LLMs have actually done so.

Load-bearing premise

The load-bearing premise is that proximity in embedding space is functionally equivalent to sharing a syntactic property, so that all semantically similar inputs are processed by one causal common factor; if close embeddings can still feed different or overlapping causal pathways, the causal-systematicity constraint is not met.

Editorial extensions

If this is right

  • Davies's original verdict that connectionist networks cannot have tacit knowledge does not carry over to transformer LLMs, because the embedding layer supplies the missing shared-property mechanism in weakened form.
  • Fact-like behavior counts as stored knowledge only when a causal common factor can be located and shown to mediate the whole semantic family, giving researchers a principled distinction between knowledge and memorization.
  • Editing methods gain a theoretical rationale: generalization and specificity scores become operational measures of Davies-style causal systematicity rather than purely practical metrics.
  • The framework gives a systematic target for correcting bias, hallucination, and misinformation: edit the causal common factor behind a cluster of related predictions instead of patching individual prompts.
  • Because the argument for the weakened syntactic-structure constraint is architectural, it predicts that newer LLMs beyond GPT-J will show the same intervention pattern if the attribution is to hold.

Reading between the lines

Editorial extensions of the paper, not claims the author makes directly.

  • The argument is architecture-general enough that the framework should transfer to any learned-embedding transformer, which makes tacit knowledge a testable empirical hypothesis for every new LLM release rather than a one-model finding.
  • A sharper prediction follows that the paper does not test: within a paraphrase cluster, edit generalization should track embedding distance, so pairs with closer embeddings should co-update more reliably; if they do not, the embedding route is not the causal carrier and the weakened constraint would need revision.
  • The Hase et al. replication caveat points to an extension: causal common factors may be distributed across layers, so multi-layer interventions could either edit tacit knowledge more robustly or reveal that the 'factor' is a network-level structure rather than a single MLP module.
  • If tacit knowledge is accepted as a causal-explanatory structure, the philosophical dispute about whether LLMs 'really' know shifts from questions of justification and conceptual content to the empirical question of whether an internal structure plays the right causal role.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, and a circularity audit.

Referee Report

3 major / 5 minor

Summary. The paper argues that transformer-based LLMs can acquire tacit knowledge in the sense defined by Martin Davies (1990), despite Davies's own claim that connectionist networks cannot satisfy the relevant constraints. The argument proceeds by endorsing Davies's constraints of semantic description and causal systematicity, while weakening the syntactic-structure constraint: instead of requiring identical input-layer patterns for semantically similar inputs, the paper proposes that the embedding layer's property of mapping semantically similar inputs to proximate regions in vector space suffices. The empirical support comes from Meng et al.'s (2022) ROME experiments, in which edits to MLP key-value pairs generalize to semantically similar prompts and show specificity, which the paper interprets as evidence that MLP modules function as Davies-style causal common factors. The paper concludes that tacit knowledge is a promising conceptual framework for describing and intervening on LLM internals, while acknowledging that the evidence is preliminary and that polysemanticity and localization concerns remain.

Significance. If the argument succeeds, the paper makes a valuable interdisciplinary contribution: it gives a precise, philosophically grounded vocabulary for attributing knowledge to LLMs and connects Davies's classic account to contemporary mechanistic interpretability. The paper is also commendably candid about its limitations, explicitly flagging the preliminary status of the evidence, Hase et al.'s replication caveats, and polysemanticity concerns. The paper's proposal that MLP key-value pairs, as revealed by causal tracing and ROME, can be interpreted as causal common factors is a concrete, potentially fruitful bridge between philosophy of cognitive science and interpretability research.

major comments (3)
  1. [Section 4, weakening the syntactic-structure constraint] The central load-bearing step is the claim that proximity in embedding space is functionally equivalent to Davies's shared syntactic property, so that semantically similar inputs are recognized and processed by a shared causal common factor. This equivalence is asserted, not demonstrated. Proximity is a static geometric property of input representations, whereas Davies's constraint requires a property that determines causal processing downstream; the paper does not show that nearby inputs are routed through the same MLP keys, attention heads, or other causal structures. For multi-word prompts, semantic similarity between paraphrases does not by itself guarantee proximity of their contextualized, sequence-level embeddings, and the cited evidence (Abdou et al. 2021; Grand et al. 2022) concerns word-level semantic properties rather than the causal processing of full inputs. Because the weakened constraint is what allows the paper to bypass Davies's own objection, this gap is load-bearing for the central claim.
  2. [Section 5.2, ROME evidence as support for causal common factors] The ROME evidence does not clearly distinguish a Davies-style semantic common factor from token-level key matching. Figure 4 and the accompanying text state that the edit is applied at the last token of the subject, and Meng et al.'s generalization tests use paraphrases that retain that subject noun phrase. Such prompts may activate the same or nearby MLP keys regardless of any general semantic categorizer at the embedding layer, so the observed generalization is consistent with a much narrower causal structure than the one required for tacit knowledge. The paper needs to address this alternative explanation explicitly, for example by discussing whether paraphrase generalization persists when the subject is lexically varied while the semantic content is preserved.
  3. [Section 5.3, implications of Hase et al. for causal localization] The paper cites Hase et al. (2023) as showing that edits at many different locations have similar efficacy, and then says this means causal tracing might not reliably localize the relevant causal common factors. This concession is in tension with the paper's earlier claim that MLP modules are the causal common factors. If equally effective edits can be made at arbitrary locations, the intervention-based evidence for a unique, semantically structured causal common factor is weakened. The paper should clarify how, given Hase et al.'s findings, the ROME results still provide evidence specifically for Davies-style causal systematicity rather than for a more distributed or redundant causal structure that does not mirror semantic patterns.
minor comments (5)
  1. [Abstract and Introduction] The abstract says the paper 'demonstrates' that architectural features of LLMs satisfy the constraints, but Section 5.3 repeatedly characterizes the evidence as preliminary and notes several caveats. The wording should be aligned with this more cautious framing, for example by saying 'argues' and 'provides preliminary evidence'.
  2. [Section 4, paragraph beginning 'Moreover, the fact...'] The phrase 'semantically similar inputs are close together in the high-dimensional vector space also guides the network towards similar processing pathways' is stated without a citation or argument; either provide supporting evidence or mark this as a hypothesis to be tested.
  3. [Section 3.3.2, Figure 3] The figure caption uses 'A)' and 'B)' inconsistently with the body text, which uses 'a' and 'b' in figure 2; standardize the caption formatting for clarity.
  4. [Footnote 4] The footnote clarifying that Davies's 'syntax' differs from linguistic syntax is helpful, but it would be useful to also note that the paper's use of 'syntactic structure' throughout refers to this Davies-specific sense, especially in Section 4 where the term is weakened.
  5. [References] The reference list appears complete, but some entries (e.g., Meng et al. 2022) lack page numbers or DOIs while others include them; standardizing the reference format would improve readability.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity found: the argument is conditional, uses external empirical evidence, and does not relabel a fitted parameter as a prediction.

full rationale

The paper's central claim is conditional: if LLMs meet Davies' constraints, then they can be attributed tacit knowledge, and the paper argues that some LLMs do meet them. The constraints are taken from Davies (1990), not constructed from the target conclusion. The weakening of the syntactic-structure constraint to embedding proximity is an explicit definitional proposal, defended by appeal to the distributional hypothesis and to architectural differences between older connectionist networks and transformer-based LLMs; it is not a hidden restatement of the conclusion. The empirical component rests on Meng et al. (2022), an external experimental study of ROME interventions, and the paper explicitly reports its own reservations from Hase et al. (2023) and from polysemanticity. No parameter is fitted and then renamed as a prediction; no uniqueness theorem from the author's prior work is invoked; and there are no self-citations carrying the argument. The closest thing to a concern is that the weakened constraint may not fully capture Davies' intended shared-syntactic-property requirement, but that is a philosophical adequacy question, not a circularity. A redefinition can be mistaken or too weak without being circular. Therefore the derivation chain is not equivalent to its inputs by construction.

Assumptions & free parameters 0 free parameters · 5 assumptions · 0 invented entities

No numerical fitting is involved; the central assumptions are conceptual: Davies's framework, the distributional hypothesis, and the paper's own weakened version of syntactic structure.

assumptions (5)
  • domain assumption Davies's three constraints are the correct conditions for tacit knowledge attribution.
    The paper endorses Davies's framework without independent defense; the central claim is defined in terms of these constraints.
  • domain assumption Word embeddings encode semantic similarity as vector proximity (distributional hypothesis).
    Section 4 relies on this to argue the weakened syntactic structure constraint is met.
  • ad hoc to paper Proximity in embedding space causes semantically similar inputs to engage a shared causal common factor.
    This is the load-bearing weakening introduced in Section 4; it is asserted rather than proven and is the main point Hase et al.'s replication concern bears on.
  • domain assumption Generalization and specificity of model edits indicate a causal common factor was targeted.
    Section 5.2 interprets Meng et al.'s metrics in Davies's causal terms; the paper itself notes this inference is preliminary.
  • domain assumption Causal tracing and ROME interventions identify causally relevant MLP modules.
    Section 5.1 imports Meng et al.'s methodology; Section 5.3 acknowledges Hase et al. (2023) found localization may not be the best target for editing.

how reviews work

0 comments
Cite this review

Pith. "Pith review of What Do Large Language Models Know? Tacit Knowledge as a Potential Causal-Explanatory Structure." pith.science (2026). https://pith.science/paper/NIABZENX

@misc{pith2026250412187,
  author       = {Pith},
  title        = {Pith review of: What Do Large Language Models Know? Tacit Knowledge as a Potential Causal-Explanatory Structure},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/NIABZENX}},
  note         = {Machine review of arXiv:2504.12187}
}
read the original abstract

It is sometimes assumed that Large Language Models (LLMs) know language, or for example that they know that Paris is the capital of France. But what -- if anything -- do LLMs actually know? In this paper, I argue that LLMs can acquire tacit knowledge as defined by Martin Davies (1990). Whereas Davies himself denies that neural networks can acquire tacit knowledge, I demonstrate that certain architectural features of LLMs satisfy the constraints of semantic description, syntactic structure, and causal systematicity. Thus, tacit knowledge may serve as a conceptual framework for describing, explaining, and intervening on LLMs and their behavior.

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

36 extracted references · 10 canonical work pages

  1. [7]

    Transparency in Complex Computational Systems

    “Transparency in Complex Computational Systems.” Philosophy of Science 87 (4): 568–89. https://doi.org/10.1086/709729. Davies, Martin

    doi:10.1086/709729

  2. [10]

    Editing Factual Knowledge in 38 Language Models

    “Editing Factual Knowledge in 38 Language Models.” arXiv. https://doi.org/10.48550/arXiv.2104.08164. Dretske, Fred I

    arXiv doi:10.48550/arxiv.2104.08164

  3. [11]

    Toy Models of Superposition

    “Toy Models of Superposition.” arXiv. https://doi.org/10.48550/arXiv.2209.10652. Evans, Gareth

    arXiv doi:10.48550/arxiv.2209.10652

  4. [15]

    Transformer Feed-Forward Layers Are Key-Value Memories

    “Transformer Feed-Forward Layers Are Key-Value Memories.” arXiv. https://doi.org/10.48550/arXiv.2012.14913. Grand, Gabriel, Idan Asher Blank, Francisco Pereira, and Evelina Fedorenko

    arXiv doi:10.48550/arxiv.2012.14913 2012

  5. [16]

    Semantic Projection Recovers Rich Human Knowledge of Multiple Object Features from Word Embeddings

    “Semantic Projection Recovers Rich Human Knowledge of Multiple Object Features from Word Embeddings.” Nature Human Behaviour 6 (7): 975–87. https://doi.org/10.1038/s41562-022-01316-8. 39 Graves, Alex

    doi:10.1038/s41562-022-01316-8

  6. [19]

    Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizing Model Beliefs

    “Do Language Models Have Beliefs? Methods for Detecting, Updating, and Visualizing Model Beliefs.” ArXiv. https://doi.org/10.48550/arXiv.2111.13654. Hinton, G. E., J. L. McClelland, and D. E. Rumelhart

    arXiv doi:10.48550/arxiv.2111.13654

  7. [20]

    Explanations in AI as Claims of Tacit Knowledge

    “Explanations in AI as Claims of Tacit Knowledge.” Minds and Machines . https://doi.org/10.1007/s11023-021-09588-1. Lapuschkin, Sebastian, Stephan Wäldchen, Alexander Binder, Grégoire Montavon, 40 Wojciech Samek, and Klaus-Robert Müller

    work page doi:10.1007/s11023-021-09588-1

  8. [22]

    Syntactic Structure from Deep Learning

    “Syntactic Structure from Deep Learning.” Annual Review of Linguistics 7 (1): 195–212. https://doi.org/10.1146/annurev-linguistics-032020-051035. Mahowald, Kyle, Anna A. Ivanova, Idan A. Blank, Nancy Kanwisher, Joshua B. Tenenbaum, and Evelina Fedorenko

    doi:10.1146/annurev-linguistics-032020-051035

Show all 36 references
  1. [25]

    Zoom In: An Introduction to Circuits

    “Zoom In: An Introduction to Circuits.” Distill 5 (3): e00024.001. https://doi.org/10.23915/distill.00024.001. OpenAI

    doi

  2. [27]

    Meaning without Reference in Large Language Models

    “Meaning without Reference in Large Language Models.” arXiv. https://doi.org/10.48550/arXiv.2208.02957. Quine, W.V

    doi arXiv

  3. [29]

    Talking About Large Language Models

    “Talking About Large Language Models.” arXiv. https://doi.org/10.48550/arXiv.2212.03551. Smolensky, Paul

    doi arXiv

  4. [30]

    On the Proper Treatment of Connectionism

    “On the Proper Treatment of Connectionism.” Behavioral and 42 Brain Sciences 11 (1): 1–23. https://doi.org/10.1017/S0140525X00052432. Sukhbaatar, Sainbayar, Edouard Grave, Guillaume Lample, Herve Jegou, and Armand Joulin

    doi

  5. [32]

    Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet

    “Scaling Monosemanticity: Extracting Interpretable Features from Claude 3 Sonnet.” Transformer Circuits Thread . https://transformer-circuits.pub/2024/scaling-monosemanticity/index.html. Van Gelder, Tim

    2024

  6. [34]

    From Task Structures to World Models: What Do LLMs Know?

    “From Task Structures to World Models: What Do LLMs Know?” Trends in Cognitive Sciences 28 (5): 404–15. https://doi.org/10.1016/j.tics.2024.02.008. Zednik, Carlos

    2024 doi

  7. [36]

    Solving the Black Box Problem: A Normative Framework for Explainable Artificial Intelligence

    “Solving the Black Box Problem: A Normative Framework for Explainable Artificial Intelligence.” Philosophy & Technology 34 (2): 265–88. https://doi.org/10.1007/s13347-019-00382-7. 44 Acknowledgements : I am grateful to Carlos Zednik, Vincent Müller, Lambèr Royakkers, and Charl...

    doi

  8. [1096]

    Lenat, Douglas B., and R

    https://doi.org/10.1038/s41467-019-08987-4. Lenat, Douglas B., and R. V. Guha

    doi

  9. [1954]

    Distributional Structure

    “Distributional Structure.” WORD 10 (2–3): 146–62. https://doi.org/10.1080/00437956.1954.11659520. Hase, Peter, Mohit Bansal, Been Kim, and Asma Ghandeharioun

    1954

  10. [1957]

    A Synopsis of Linguistic Theory, 1930-1955

    “A Synopsis of Linguistic Theory, 1930-1955.” In Studies in Linguistic Analysis , edited by John Rupert Firth, 1–32. Oxford: Blackwell. Fodor, Jerry A., and Zenon W. Pylyshyn

    1930

  11. [1970]

    Methodological Reflections on Current Linguistic Theory

    “Methodological Reflections on Current Linguistic Theory.” Synthese 21:13. https://doi.org/10.1007/978-94-010-2557-7_14. Shanahan, Murray

    doi

  12. [1981]

    Psychologism and Behaviorism

    “Psychologism and Behaviorism.” The Philosophical Review 90 (1): 5–43. https://doi.org/10.2307/2184371. Buckner, Cameron J

    doi

  13. [1988]

    Connectionism and Cognitive Architecture: A Critical Analysis

    “Connectionism and Cognitive Architecture: A Critical Analysis.” Cognition 28 (1): 3–71. https://doi.org/10.1016/0010-0277(88)90031-5. Gascoigne, Neil, and Tim Thornton

    doi

  14. [1990]

    Knowledge of Rules in Connectionist Networks

    “Knowledge of Rules in Connectionist Networks.” Intellectica. Revue de l’Association Pour La Recherche Cognitive 9 (1): 81–126. https://doi.org/10.3406/intel.1990.881. Davies, Martin

    1990 doi

  15. [1991]

    Systematicity, Structured Representations and Cognitive Architecture: A Reply to Fodor and Pylyshyn

    “Systematicity, Structured Representations and Cognitive Architecture: A Reply to Fodor and Pylyshyn.” In Connectionism and the Philosophy of Mind , edited by Terence Horgan and John Tienson, 198–218. Dordrecht: Springer Netherlands. https://doi.org/10.1007/978-94-011-3524-5_9...

    doi

  16. [1992]

    Defining ‘Distributed Representation.’

    “Defining ‘Distributed Representation.’” Connection Science 4 (3–4): 175–91. https://doi.org/10.1080/09540099208946614. Vaswani, Ashish, Noam Shazeer, Niki Parmar, Jakob Uszkoreit, Llion Jones, Aidan N Gomez, Łukasz Kaiser, and Illia Polosukhin

    doi

  17. [2011]

    Natural Language Processing: An Introduction

    “Natural Language Processing: An Introduction.” Journal of the American Medical Informatics Association 18 (5): 544–51. https://doi.org/10.1136/amiajnl-2011-000464. Olah, Chris, Nick Cammarata, Ludwig Schubert, Gabriel Goh, Michael Petrov, and Shan Carter

    2011 doi

  18. [2012]

    Long Short-Term Memory

    “Long Short-Term Memory.” In Supervised Sequence Labelling with Recurrent Neural Networks , edited by Alex Graves, 37–45. Studies in Computational Intelligence. Berlin, Heidelberg: Springer. https://doi.org/10.1007/978-3-642-24797-2_4. Harris, Zellig S

    doi

  19. [2014]

    London: Routledge

    Tacit Knowledge . London: Routledge. https://doi.org/10.4324/9781315729886. Geva, Mor, Roei Schuster, Jonathan Berant, and Omer Levy

    doi

  20. [2015]

    Knowledge (Explicit, Implicit and Tacit): Philosophical Aspects

    “Knowledge (Explicit, Implicit and Tacit): Philosophical Aspects.” In International Encyclopedia of the Social & Behavioral Sciences , 74–90. Elsevier. https://doi.org/10.1016/B978-0-08-097086-8.63043-X. De Cao, Nicola, Wilker Aziz, and Ivan Titov

    doi

  21. [2016]

    How the Machine ‘Thinks’: Understanding Opacity in Machine Learning Algorithms

    “How the Machine ‘Thinks’: Understanding Opacity in Machine Learning Algorithms.” Big Data & Society 3 (1): 205395171562251. https://doi.org/10.1177/2053951715622512. 37 Chang, Tyler, Zhuowen Tu, and Benjamin Bergen

    doi

  22. [2018]

    Will Machine Learning Yield Machine Intelligence?

    “Will Machine Learning Yield Machine Intelligence?” In Philosophy and Theory of Artificial Intelligence 2017 , edited by Vincent Müller, 44:225–27. Studies in Applied Philosophy, Epistemology and Rational Ethics. Springer International Publishing. https://doi.org/10.1007/978-3...

    2017 doi

  23. [2019]

    Augmenting Self-Attention with Persistent Memory

    “Augmenting Self-Attention with Persistent Memory.” arXiv. https://doi.org/10.48550/arXiv.1907.01470. Templeton, Adly, Tom Conerly, Jonathan Marcus, Jack Lindsey, Trenton Bricken, Brian Chen, Adam Pearce, et al

    1907 doi arXiv

  24. [2020]

    Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data

    “Climbing towards NLU: On Meaning, Form, and Understanding in the Age of Data.” In Proceedings of the 58th Annual Meeting of the Association for Computational Linguistics , 5185–98. Online: Association for Computational Linguistics. https://doi.org/10.18653/v1/2020.acl-main.46...

    doi

  25. [2021]

    Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Color

    “Can Language Models Encode Perceptual Structure Without Grounding? A Case Study in Color.” arXiv. https://doi.org/10.48550/arXiv.2109.06129. Bahdanau, Dzmitry, Kyung Hyun Cho, and Yoshua Bengio

    doi arXiv

  26. [2022]

    The Geometry of Multilingual Language Model Representations

    “The Geometry of Multilingual Language Model Representations.” In Proceedings of the 2022 Conference on Empirical Methods in Natural Language Processing , 119–36. Abu Dhabi, United Arab Emirates: Association for Computational Linguistics. https://doi.org/10.18653/v1/2022.emnlp...

    2022 doi

  27. [2023]

    Symbols and Grounding in Large Language Models

    “Symbols and Grounding in Large Language Models.” Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences 381 (2251): 20220041. https://doi.org/10.1098/rsta.2022.0041. Piantadosi, Steven T., and Felix Hill

    2022

  28. [2024]

    Dissociating Language and Thought in Large Language Models

    “Dissociating Language and Thought in Large Language Models.” Trends in Cognitive Sciences 28 (6): 517–40. https://doi.org/10.1016/j.tics.2024.01.011. Meng, Kevin, David Bau, Alex Andonian, and Yonatan Belinkov

    2024 doi

Pith tools

Reviewed August 16, 2026 · model on record in the stance chip above.