Pith. sign in

REVIEW 4 major objections 6 minor 97 references

A quantum semantic framework for natural language processing

T0 review · 4 major / 6 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read The paper claims that meaning is not stored in text but actualized by an interpreter, and LLM-based Bell tests with several CHSH values above the classical bound support a quantum-like account of ambiguous language.

desk verdict The Bell test on LLMs is a real idea, but the evidence for contextuality collapses once you take the paper's own no-signaling caveat seriously. read the letter →

arxiv 2506.10077 v2 pith:MCS2BLOP submitted 2025-06-11 cs.CL cs.AIcs.IRcs.ITmath.IT

Christopher J. Agostino , Quan Le Thien , Molly Apsel , Denizhan Pak , Elina Lesyk , Ashabari Majumdar This is my paper
classification cs.CLcs.AIcs.IRcs.ITmath.IT
keywords semanticdegeneracyquantumsemanticsCHSHinequalityBelltestlargelanguagemodelsKolmogorovcomplexitycontextualityambiguity
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

Semantic degeneracy, the paper argues, is not a nuisance but a fundamental property of language: the number of possible readings of an expression grows so fast with complexity that no finite system can reliably recover one intended meaning. The authors formalize this with a Kolmogorov-complexity lower bound on the bits needed to pin down a meaning, and conclude that text cannot carry intrinsic meaning. Instead, they propose that meaning is actualized by an interpreter, and that a quantum-like formalism, with expressions as superpositions and interpretations as measurements, is the natural description. To test this, they run a CHSH Bell test with LLM readers and report correlation values above the classical bound in several runs. The payoff is a reorientation of NLP away from single correct answers toward Bayesian-style sampling of an interpretation space.

What carries the argument

The load-bearing machinery is the CHSH semantic Bell test. Two LLM-based agents, Alice and Bob, each read a sentence containing two ambiguous words under two priming settings ($A,A'$ and $B,B'$); a separate LLM call classifies each agent's free-text interpretation into one of two meanings per word, coded $\pm1$, and the four pair-correlations are combined in $S=E(A,B)-E(A,B')+E(A',B)+E(A',B')$. Classical realistic theories require $|S|\le 2$, and the paper reads $|S|>2$ as non-classical contextuality. The theoretical side is the Kolmogorov-complexity bound $K(M(S_E)) \ge \sum_i c_i + \sum_{i,j} c_{i,j}^2$ on the information needed to fix one intended meaning, plus the quantum formalism in which $S_E$ is a state $|\psi_{S_E}\rangle$ in a Hilbert space and the interpreter applies a Hermitian observable $\hat{\mu}$ whose eigenbasis is agent- and context-dependent.

What would settle it

Run the same protocol with unambiguous control words and with human raters or a forced-choice interface replacing the second LLM classification; if $|S|$ drops to 2 or below in either case, the apparent contextuality is an artifact of the measurement pipeline. Also apply the contextuality-by-default correction for inconsistent connectedness: the violation must survive that adjustment to support true non-classical contextuality rather than direct contextual influence.

Watch

Extended reading notes

Core claim

The paper's central claim is that an act of linguistic interpretation under ambiguity can be non-classically contextual: the correlation pattern across interpretive settings can violate the CHSH bound $|S|\le 2$, which a classical realistic account cannot produce. In the reported experiments, eight LLM-based runs with varied personas and ambiguous word pairs produced average $|S|$ values from 1.2 to 2.8, with several runs exceeding the classical bound (2.32, 2.33, 2.44, and one at 2.8). The authors interpret this as evidence that meaning is not a pre-existing property of text: the expression is a superposition of possible readings, and interpretation is a measurement that actualizes one of them. On this view, different interpretive observables can fail to commute, so the order or context of probing changes the realized meaning rather than revealing a fixed one.

Load-bearing premise

The result stands or falls on treating an LLM's free-text paraphrase, after a second LLM maps it to a plus/minus code in each setting, as a clean Bell measurement, even though the paper concedes that the usual no-signaling condition is violated, which can by itself push a correlation above the classical bound.

Editorial extensions

If this is right

  • If semantic degeneracy is fundamental, LLM plateaus on complex, ambiguous tasks cannot be cured by more data or compute alone; the bottleneck is the combinatorial growth of required context.
  • Single-output systems are the wrong target: practical NLP should return distributions over interpretations obtained by repeated sampling under varied contexts.
  • The CHSH violations imply that classical frequentist evaluation of language systems is lossy, and Bayesian characterizations are more appropriate.
  • Observer-dependent meaning makes human-in-the-loop validation a permanent architectural feature rather than a stopgap.
  • Quantum-inspired models become a candidate formal language for order effects and non-commutative interpretations in NLP.

Reading between the lines

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

  • Beyond the paper: a direct control condition, using unambiguous sentences or words with only one sense, should give CHSH values at or below 2; if it does not, the violation is not tied to ambiguity.
  • Beyond the paper: replacing the second LLM classification with human judges or a forced-choice interface would test whether the contextuality survives the measurement pipeline rather than arising from the model's self-report.
  • Beyond the paper: the Kolmogorov-complexity argument implies that even repeated Bayesian sampling cannot converge to a unique meaning for sufficiently complex text; the practical goal becomes calibrated distributions over interpretations, not recovery of the single intended one.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

4 major / 6 minor

Summary. The paper argues that semantic degeneracy—the combinatorial explosion of possible interpretations—makes the recovery of a single intended meaning computationally intractable for large language models, proposes a quantum semantic formalism in which meaning is actualized by an observer's interpretive act, and reports a CHSH-style semantic Bell test using LLM agents. The authors interpret several measured |S|>2 values as evidence of non-classical contextuality in linguistic interpretation and conclude that repeated Bayesian-style sampling is a more appropriate methodology than single-shot interpretation.

Significance. If the central empirical claim were sound, this would be a striking result: it would provide evidence that natural language interpretation, as instantiated in LLMs, exhibits non-classical contextuality, with broad implications for NLP, cognitive science, and the philosophy of meaning. The paper addresses an important and timely problem—the reliability of LLMs on ambiguous text—and the use of diverse LLM agents as interpretive probes is creative. The authors also provide a link to their open-source npcpy package, which is a step toward reproducibility, and they explicitly acknowledge the known issue of inconsistent connectedness in cognitive Bell tests. However, as argued below, the central claim is not currently supported: the reported CHSH violations are not a valid test of contextuality under the paper's own acknowledged violation of no-signaling, the measurement pipeline is unvalidated, the sample sizes are too small for the claimed significance, and the theoretical derivation in Section II is asserted rather than derived.

major comments (4)
  1. [Section IV.B, Section V, Conclusion 2] The paper acknowledges that the CHSH no-signaling assumption is often violated in cognitive and behavioral experiments and that 'inconsistent connectedness' can produce |S|>2 without true contextuality, yet Section V and Conclusion 2 treat every |S|>2 as direct evidence of non-classical contextuality. Because the CHSH bound applies only under marginal consistency, and because the experimental design explicitly varies personas and contexts between settings, the reported violations are compatible with classical non-contextual explanations unless marginal consistency is tested directly. No such test, nor an alternative analysis such as Contextuality-by-Default, is reported; therefore the central empirical claim is not established.
  2. [Section IV.B] The outcome measurement uses a second, unvalidated LLM call to map each free-text interpretation to +1/-1, with re-interpretation for unclear or out-of-scope answers. This classifier is part of the measurement apparatus; if it is biased or unstable, the computed E(XY) correlations and S values reflect the classifier's behavior rather than the semantic interpretation itself. The paper reports no reliability statistics for this mapping, such as agreement with human coding, test-retest consistency, or confusion matrices.
  3. [Table I and Section V] The empirical demonstration consists of eight runs with N between 5 and 200 trials, with no confidence intervals, p-values, or effect sizes. Four of the eight runs yield |S|<=2, and the highest violation (|S|=2.8) comes from a five-trial run. Without a statistical model of the sampling distribution of S (e.g., bootstrap or permutation tests), the claims of 'frequent and significant violations' are unsupported; with N=5, even independent random responses can produce |S| values near 2 by chance.
  4. [Section II, Eqs. (1)-(3)] The claimed superlinear lower bound on K(M(SE)) is not derived from Kolmogorov complexity; it assumes a specific decomposition into independent concept bits c_i and pairwise bits c_{i,j}, both of which are free parameters. Equation (2) introduces an unexplained factor 1/N!, with N undefined at that point, and assumes independent per-bit error probabilities, so Eq. (3) simply restates that assumption in exponential form. Thus the 'demonstration' of exponentially vanishing probability of perfect interpretation is an assumed model, not a theorem about semantic degeneracy.
minor comments (6)
  1. [Eq. (1)] The notation in Eq. (1) is malformed: the expression 'NX i,j · c2 i,j' should be written as a consistent double sum, and the relationship between c_{i,j} (described earlier as bits) and c^2_{i,j} is unexplained.
  2. [Abstract] The abstract says 'average CHSH expectation values from 1.2 to 2.8,' but Table I reports per-run |S| values, not averages; this wording should be clarified.
  3. [Section V, Figure 2] The phrase 'the evolution of the |S| value's evolution' is redundant, and the figure would benefit from error bars or a display of trial-to-trial variability.
  4. [Reference [1]] Reference [1] contains a typo: 'What Computers Still Can?T Do' should be 'What Computers Still Can't Do.'
  5. [Footnote 1] The speculation about Claude's 'French mindset' affecting its effort based on time of year is unsupported and unrelated to the experiment; it should be removed or replaced with empirical evidence.
  6. [Section VII, Conclusion 5] Conclusion 5 claims that the observed statistical properties are 'objective, structural features of natural language itself,' which overgeneralizes from LLM agent responses to language as such; this inference is not justified by the data.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the formal derivations are explicit model assumptions, and the Bell-test conclusion is an empirical inference rather than an algebraic reduction to inputs.

full rationale

The theoretical part (Section II) is not circular: Eq. (3) is a direct mathematical consequence of the stated independence and per-bit-degeneracy assumptions in Eq. (2), and the paper does not fit Eq. (3) to data and then present that fit as a prediction. Section III's quantum formalism is explicitly posited ("We begin by positing..."), and the CHSH experiment in Sections IV–V is an independent empirical probe of the framework's non-classical hypothesis rather than a restatement of the framework's definitions. No fitted quantum parameter is recycled into the S-value computation. There is also no load-bearing self-citation chain: the paper's citations to prior quantum-cognition work are to external authors, and no uniqueness theorem is imported from the authors' own prior work. The most serious weakness is the acknowledged no-signaling caveat in Section IV.B, where the authors state that "inconsistent connectedness" can produce |S|>2 without true contextuality, and then interpret |S|>2 as evidence of contextuality; the separate LLM classifier is also unvalidated. These are validity threats, not circularity in the sense of a conclusion being equivalent to its premises by construction. Accordingly, no circular steps are identified and the circularity score is 0.

Assumptions & free parameters 3 free parameters · 5 assumptions · 3 invented entities

The framework imports several unmeasured quantities and assumptions; the core empirical conclusion about non-classical contextuality rests on axioms 3 and 4, which are not justified by the reported data.

free parameters (3)
  • c_i: bits to instantiate each concept = 5 bits per concept in Figure 1
    Eq (1) assumes a fixed cost per concept; Figure 1 uses 5 bits for illustration, but no measurement supports this.
  • c_{i,j}: bits for pairwise concept relationships = 1 bit per relationship in Figure 1
    Eq (1) introduces a quadratic term and Figure 1 uses 1 bit; the quadratic growth is assumed, not derived.
  • average degeneracy per bit \bar{d}_b = not specified (varied as p_e in Figure 1)
    Eq (3) needs an average per-bit degeneracy to produce the exponential decay; the value is assumed and no data are given.
assumptions (5)
  • domain assumption The intended meaning M(SE) has a well-defined Kolmogorov complexity K(M(SE)) and the program that specifies it exists.
    Section II defines K(M(SE)) but gives no encoding scheme for meanings; this is a conceptual leap from strings to meanings.
  • ad hoc to paper The total information to specify meaning decomposes into independent bits with independent per-bit error probabilities.
    Eqs (2)-(3) multiply independent probabilities, which is what creates the exponential decay; no empirical or theoretical support is provided.
  • domain assumption LLM agents with persona prompts can serve as observers analogous to humans in a cognitive Bell test.
    Section IV.A argues LLMs 'collapse' interpretations like brains, but cites no evidence that LLM response distributions satisfy the same probabilistic constraints as human judgments.
  • ad hoc to paper The CHSH test remains valid for semantic interpretation data even when marginal consistency is violated.
    Section IV.B cites Dzhafarov's warning that inconsistent connectedness can produce |S|>2, but the conclusions still interpret high |S| as non-classical contextuality.
  • domain assumption Semantic expressions can be represented as vectors in a Hilbert space with unknown basis states, and interpretation is a quantum measurement.
    Section III postulates |psi_SE> and the observable \hat{mo} without any operational definition of the basis or the observable.
invented entities (3)
  • Semantic state vector |psi_SE> in a semantic Hilbert space H_S
    purpose: To represent the superposition of possible interpretations of an expression.
    No independent prediction is made outside the Bell experiment; the only evidence is the contested experimental result.
  • Interpretive observable \hat{mo} (Japanese 'mo')
    purpose: To model an agent's interpretation as a quantum measurement.
    No specific operator is constructed; it is defined only by its role in the formalism.
  • Semantic Hamiltonian \hat{H_sem}(t) and scale constant \hbar_sem
    purpose: To govern time evolution of semantic states under Eq (5).
    The authors explicitly set this component aside and do not use it; no value or observable consequence is given.

how reviews work

0 comments
Cite this review

Pith. "Pith review of A quantum semantic framework for natural language processing." pith.science (2026). https://pith.science/paper/MCS2BLOP

@misc{pith2026250610077,
  author       = {Pith},
  title        = {Pith review of: A quantum semantic framework for natural language processing},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MCS2BLOP}},
  note         = {Machine review of arXiv:2506.10077}
}
abstract

Semantic degeneracy represents a fundamental property of natural language that extends beyond simple polysemy to encompass the combinatorial explosion of potential interpretations that emerges as semantic expressions increase in complexity. In this work, we argue this property imposes fundamental limitations on Large Language Models (LLMs) and other modern NLP systems, precisely because they operate within natural language itself. Using Kolmogorov complexity, we demonstrate that as an expression's complexity grows, the amount of contextual information required to reliably resolve its ambiguity explodes combinatorially. The computational intractability of recovering a single intended meaning for complex or ambiguous text therefore suggests that the classical view that linguistic forms possess intrinsic meaning in and of themselves is conceptually inadequate. We argue instead that meaning is dynamically actualized through an observer-dependent interpretive act, a process whose non-deterministic nature is most appropriately described by a non-classical, quantum-like logic. To test this hypothesis, we conducted a semantic Bell inequality test using diverse LLM agents. Our experiments yielded average CHSH expectation values from 1.2 to 2.8, with several runs producing values (e.g., 2.3-2.4) in significant violation of the classical boundary ($|S|\leq2$), demonstrating that linguistic interpretation under ambiguity can exhibit non-classical contextuality, consistent with results from human cognition experiments. These results inherently imply that classical frequentist-based analytical approaches for natural language are necessarily lossy. Instead, we propose that Bayesian-style repeated sampling approaches can provide more practically useful and appropriate characterizations of linguistic meaning in context.

Figures

Figures reproduced from arXiv: 2506.10077 by the authors.

Figure 1
Figure 1. FIG. 1. Probability of perfect semantic interpretation versus [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. FIG. 2. Example evolution of the calculated CHSH S-value [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗

Discussion (0). Sign in to comment.

Reference graph

Works this paper leans on

97 extracted references · 63 canonical work pages

  1. [1]

    DSM-powered approaches still exhibit limitations particularly when dealing with complex, ambigu- ous, or context-rich texts [11, 12]

  2. [2]

    small worlds

    The apparent lack of much progress seen in the latest generation of frontier large language models (LLMs)—compared tothe leaps seen inprior model releases—appears to have revealed a fundamental barrier in the reliability of semantic problem solv- ing that more compute/data have not alleviated (e.g., GPT-4.5, Llama 4, Gemini 2.5, Claude 3.7 Sonnet). Likewi...

  3. [3]

    computational cognitive systems

    It is interesting then that one of our experi- ments(albeitN= 5), the|S|valuereached2.8, providing an empirical upper bound for these experiments that can be consistent with quantum computational frameworks. VI. DISCUSSION The arguments presented in this work, concerning the fundamental limitations imposed by semantic degener- acy and the potential of a q...

  4. [4]

    Semantic degeneracy is a fundamental property of natural language that imposes an information- theoretic limit on interpretation; our analysis using Kolmogorov complexity (Section II) formalizes how this makes the recovery of a single intended mean- ing from a complex expression computationally in- tractable for any system, thereby providing a clear expla...

  5. [5]

    Linguistic interpretation under ambiguity exhibits non-classical contextuality, as demonstrated by fre- quent and significant violations of the CHSH in- equality (|S|>2) in our semantic Bell test experi- ments with LLM agents (Sections IV, V)

  6. [6]

    The contextuality measured in the interpretive acts of LLMs is consistent with a broader pattern of non-classical findings across human cognitive sci- ence, indicating that observer-dependence and in- determinacy are general principles of information processing and not simply artifacts of human psy- chology

  7. [7]

    The observer-dependent nature of meaning, con- firmed by our experiment, reveals that there is no absolute, fundamental meaning to be found, only contextualized interpretations; consequently, the only viable scientific methodology is to shift from seeking any single “correct” answer and in- stead use repeated Bayesian-style sampling to char- acterizehowth...

  8. [8]

    The consistent emergence of non-classical contex- tuality across a diverse pool of non-biological LLM agents, when considered alongside similar findings in human cognition, indicates that these statistical properties are not artifacts of any specific interpre- tive system but are objective, structural features of natural language itself

Show all 97 references
  1. [9]

    H. L. Dreyfus,What Computers Still Can?T Do: A Cri- tique of Artificial Reason(MIT Press, 1992)

    1992

  2. [10]

    Z. S. Harris, Distributional Structure, WORD10, 146 (1954)

    1954

  3. [11]

    S.C.Deerwester, S.T.Dumais, G.W.Furnas, T.K.Lan- dauer, and R. A. Harshman, Indexing by latent semantic analysis, Journal of the American Society for Information Science41, 391 (1990)

    1990

  4. [12]

    T. K. Landauer and S. T. Dumais, A Solution to Plato’s Problem: The Latent Semantic Analysis Theory of Ac- quisition, Induction, and Representation of Knowledge, Psychological Review104, 211 (1997). 10

    1997

  5. [13]

    T. L. Griffiths, M. Steyvers, and J.B. Tenenbaum,Topics in semantic representation, Psychological Review114, 211 (2007)

    2007

  6. [14]

    Mikolov, I

    T. Mikolov, I. Sutskever, K. Chen, G. S. Corrado, and J. Dean, Distributed Representations of Words and Phrases and their Compositionality, arXiv preprint arXiv:1310.4546 (2013), arXiv:1310.4546 [cs.CL]

    2013 arXiv

  7. [15]

    J.Devlin, M.-W.Chang, K.Lee,andK.Toutanova,Bert: Pre-training of deep bidirectional transformers for lan- guage understanding, arXiv Preprint arXiv:1810.04805 (2018), arXiv:1810.04805

    2018 arXiv

  8. [16]

    Radford, J

    A. Radford, J. Wu, R. Child, D. Luan, D. Amodei, and I. Sutskever, Language models are unsupervised multi- task learners, OpenAI Blog (2019), openAI technical re- port, Version 1

    2019

  9. [17]

    Brown, B

    T. Brown, B. Mann, N. Ryder, M. Subbiah, J. D. Ka- plan, P. Dhariwal, A. Neelakantan, P. Shyam, G. Sastry, A. Askell,et al., Language models are few-shot learners, inAdvances in Neural Information Processing Systems 33 (NeurIPS 2020), edited by H. Larochelle, M. Ranzato, R. Ha...

    2020

  10. [18]

    Vaswani, N

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, L. Kaiser, and I. Polosukhin, At- tention Is All You Need, arXiv preprint arXiv:1706.03762 (2017), arXiv:1706.03762 [cs.CL]

    2017 arXiv

  11. [19]

    M. N. Jones, When does abstraction occur in semantic memory: Insights from distributional models, Language, Cognition and Neuroscience34, 1338 (2018)

    2018

  12. [20]

    Hoffman, M

    P. Hoffman, M. A. Lambon Ralph, and T. T. Rogers, Se- mantic diversity: A measure of semantic ambiguity based on variability in the contextual usage of words, Behavior Research Methods45, 718 (2013)

    2013

  13. [21]

    Rabbat, Y

    M.Assran, Q.Duval, I.Misra, P.Bojanowski, P.Vincent, M. Rabbat, Y. LeCun, and N. Ballas, Self-supervised learning from images with a joint-embedding predic- tive architecture, inProceedings of the IEEE/CVF Con- ference on Computer Vision and Pattern Recognition (CVPR)(2023) pp...

    2023

  14. [22]

    Gu and T

    A. Gu and T. Dao, Mamba: Linear-Time Sequence Modeling with Selective State Spaces, arXiv e-prints , arXiv:2312.00752 (2023), arXiv:2312.00752 [cs.LG]

    2023 arXiv

  15. [23]

    Darlow, C

    L. Darlow, C. Regan, S. Risi, J. Seely, and L. Jones, Continuous Thought Machines, arXiv e-prints , arXiv:2505.05522 (2025), arXiv:2505.05522 [stat.ML]

    2025

  16. [24]

    Q. Sun, E. Cetin, and Y. Tang, Transformer-Squared: Self-adaptive LLMs, arXiv e-prints , arXiv:2501.06252 (2025), arXiv:2501.06252 [cs.LG]

    2025 arXiv

  17. [25]

    Simonds and A

    T. Simonds and A. Yoshiyama, LADDER: Self- Improving LLMs Through Recursive Problem Decom- position, arXiv e-prints , arXiv:2503.00735 (2025), arXiv:2503.00735 [cs.LG]

    2025 arXiv

  18. [26]

    Tabossi, L

    P. Tabossi, L. Colombo, and R. Job, Accessing lexical ambiguity: Effects of context and dominance, Psycho- logical Research49, 161 (1987)

    1987

  19. [27]

    K. S. Binder and K. Rayner, Contextual strength does not modulate the subordinate bias effect: Evidence from eye fixations and self-paced reading, Psychonomic Bul- letin & Review5, 271 (1998)

    1998

  20. [28]

    Rayner, A

    K. Rayner, A. E. Cook, B. J. Juhasz, and L. Frazier, Immediate disambiguation of lexically ambiguous words during reading: Evidence from eye movements, British Journal of Psychology97, 467 (2006)

    2006

  21. [29]

    S. A. McDonald and R. C. Shillcock, Rethinking the Word Frequency Effect: The Neglected Role of Distribu- tional Information in Lexical Processing, Language and Speech44, 295 (2001)

    2001

  22. [30]

    L. W. Barsalou, Situated simulation in the human con- ceptual system, Language and Cognitive Processes18, 513 (2003)

    2003

  23. [31]

    Yeh and L

    W. Yeh and L. W. Barsalou, The Situated Nature of Concepts, The American Journal of Psychology119, 349 (2006)

    2006

  24. [32]

    E. T. Higgins, Knowledge activation: Accessibility, ap- plicability, and salience, inSocial Psychology: Handbook of Basic Principles, edited by E. T. Higgins and A. W. Kruglanski (Guilford Press, 1996) pp. 133–168

    1996

  25. [33]

    K.Bicknell, J.L.Elman, M.Hare, K.McRae,andM.Ku- tas, Effects of event knowledge in processing verbal argu- ments, Journal of Memory and Language63, 489 (2010)

    2010

  26. [34]

    Kintsch, Predication, Cognitive Science25, 173 (2001)

    W. Kintsch, Predication, Cognitive Science25, 173 (2001)

    2001

  27. [35]

    L. A. M. Lebois, C. D. Wilson-Mendenhall, and L. W. Barsalou, Are Automatic Conceptual Cores the Gold Standard of Semantic Processing? The Context- DependenceofSpatialMeaninginGroundedCongruency Effects, Cognitive Science39, 1764 (2015)

    2015

  28. [36]

    E. Yee, S. Z. Ahmed, and S. L. Thompson-Schill, Color- less green ideas (can) prime furiously, Psychological Sci- ence23, 364 (2012)

    2012

  29. [37]

    Hoenig, E.-J

    K. Hoenig, E.-J. Sim, V. Bochev, B. Herrnberger, and M. Kiefer, Conceptual Flexibility in the Human Brain: Dynamic Recruitment of Semantic Maps from Visual, Motor, and Motion-related Areas, Journal of Cognitive Neuroscience20, 1799 (2008)

    2008

  30. [38]

    Van Dantzig, D

    S. Van Dantzig, D. Pecher, R. Zeelenberg, and L. W. Barsalou, Perceptual Processing Affects Conceptual Pro- cessing, Cognitive Science32, 579 (2008)

    2008

  31. [39]

    Bermeitinger, D

    C. Bermeitinger, D. Wentura, and C. Frings, How to switch on and switch off semantic priming effects: Ac- tivation processes in category memory depend on focus- ing specific feature dimensions, Psychonomic Bulletin & Review18, 579 (2011)

    2011

  32. [40]

    N.S.Hsu, D.J.M.Kraemer, R.T.Oliver, M.L.Schlicht- ing, and S. L. Thompson-Schill, Color, Context, and Cog- nitive Style: Variations in Color Knowledge Retrieval as aFunctionofTaskandSubjectVariables,JournalofCog- nitive Neuroscience23, 2544 (2011)

    2011

  33. [41]

    Athanasopoulos, Cognitive representation of colour in bilinguals: The case of Greek blues, Bilingualism: Lan- guage and Cognition12, 83 (2009)

    P. Athanasopoulos, Cognitive representation of colour in bilinguals: The case of Greek blues, Bilingualism: Lan- guage and Cognition12, 83 (2009)

    2009

  34. [42]

    Thompson, S

    B. Thompson, S. G. Roberts, and G. Lupyan, Cultural influences on word meanings revealed through large-scale semantic alignment, Nature Human Behaviour4, 1029 (2020)

    2020

  35. [43]

    B. T. Johns, Distributional social semantics: Inferring word meanings from communication patterns, Cognitive Psychology131, 101441 (2021)

    2021

  36. [44]

    Cesario, J

    J. Cesario, J. E. Plaks, N. Hagiwara, C. D. Navarrete, and E. T. Higgins, The Ecology of Automaticity: How Situational Contingencies Shape Action Semantics and Social Behavior, Psychological Science21, 1311 (2010)

    2010

  37. [45]

    Marian and M

    V. Marian and M. Kaushanskaya, Language context guides memory content, Psychonomic Bulletin & Review 14, 925 (2007)

    2007

  38. [46]

    L. W. Barsalou, Simulation, situated conceptualization, and prediction, Philosophical Transactions of the Royal Society B: Biological Sciences364, 1281 (2009). 11

    2009

  39. [47]

    Pecher, R

    D. Pecher, R. Zeelenberg, and J. G. W. Raaijmakers, Does Pizza Prime Coin? Perceptual Priming in Lexi- cal Decision and Pronunciation, Journal of Memory and Language38, 401 (1998)

    1998

  40. [48]

    A. M. Borghi, A. M. Glenberg, and M. P. Kaschak, Putting words in perspective, Memory & Cognition32, 863 (2004)

    2004

  41. [49]

    Connell and D

    L. Connell and D. Lynott, Principles of Representation: WhyYouCan’tRepresenttheSameConceptTwice,Top- ics in Cognitive Science6, 390 (2014)

    2014

  42. [50]

    Musz and S

    E. Musz and S. L. Thompson-Schill, Semantic variability predicts neural variability of object concepts, Neuropsy- chologia76, 41 (2015)

    2015

  43. [51]

    Vervaeke and L

    J. Vervaeke and L. Ferraro, Relevance, meaning and the cognitive science of wisdom (2013)

    2013

  44. [52]

    B. P. Andersen, M. Miller, and J. Vervaeke, Predictive processing and relevance realization: Exploring conver- gent solutions to the frame problem, Phenomenology and the Cognitive Sciences , 1 (forthcoming)

  45. [53]

    Jaeger, A

    J. Jaeger, A. Riedl, A. Djedovic, J. Vervaeke, and D. Walsh, Naturalizing relevance realization: Why agency and cognition are fundamentally not computa- tional (2023)

    2023

  46. [54]

    Vervaeke and L

    J. Vervaeke and L. Ferraro, Relevance realization and the neurodynamics and neuroconnectivity of general intelli- gence, inSmartData, edited by I. Harvey, A. Cavoukian, G. Tomko, D. Borrett, H. Kwan, and D. Hatzinakos (Springer New York, New York, NY, 2013) pp. 57–68

    2013

  47. [55]

    D. L. Hintzman, MINERVA 2: A simulation model of hu- man memory, Behavior Research Methods, Instruments, & Computers16, 96 (1984)

    1984

  48. [56]

    R. K. Jamieson, J. E. Avery, B. T. Johns, and M. N. Jones, An Instance Theory of Semantic Memory, Com- putational Brain & Behavior1, 119 (2018)

    2018

  49. [57]

    W. O. Van Dam, S.-A. Rueschemeyer, O. Lindemann, and H. Bekkering, Context Effects in Embodied Lexical- Semantic Processing, Frontiers in Psychology1, 150 (2010)

    2010

  50. [58]

    W. O. Van Dam, M. Van Dijk, H. Bekkering, and S.-A. Rueschemeyer, Flexibility in embodied lexical-semantic representations, Human Brain Mapping33, 2322 (2012)

    2012

  51. [59]

    Aerts, Quantum structure in cognition, Journal of Mathematical Psychology53, 314 (2009), special Issue: Quantum Cognition

    D. Aerts, Quantum structure in cognition, Journal of Mathematical Psychology53, 314 (2009), special Issue: Quantum Cognition

    2009

  52. [60]

    Bruza, J

    P. Bruza, J. R. Busemeyer, and L. Gabora, Introduc- tion to the special issue on quantum cognition, Journal of Mathematical Psychology53, 303 (2009), special Is- sue: Quantum Cognition

    2009

  53. [61]

    Gabora and D

    L. Gabora and D. A. and, Contextualizing con- cepts using a mathematical generalization of the quantum formalism, Journal of Experimental & Theoretical Artificial Intelligence14, 327 (2002), https://doi.org/10.1080/09528130210162253

    2002 doi

  54. [62]

    Bruza, K

    P. Bruza, K. Kitto, B. Ramm, L. Sitbon, and D. Song, Quantum-like non-separability of concept combinations, emergent associates and abduction, Logic Journal of the IGPL20, 445 (2012)

    2012

  55. [63]

    Bruza, K

    P. Bruza, K. Kitto, B. Ramm, and L. Sitbon, A proba- bilistic framework for analysing the compositionality of conceptual combinations, Journal of Mathematical Psy- chology67, 26 (2015)

    2015

  56. [64]

    J. M. Yearsley and J. R. Busemeyer, Quantum cognition and decision theories: A tutorial, Journal of Mathemati- cal Psychology74, 99 (2016), foundations of Probability Theory in Psychology and Beyond

    2016

  57. [65]

    E. M. Pothos and J. R. Busemeyer, Quantum cognition, Annual Review of Psychology73, 749 (2022)

    2022

  58. [66]

    Aerts, M

    D. Aerts, M. Czachor, B. D’Hooghe, S. Sozzo,et al., The pet-fish problem on the world-wide web., inAAAI Fall Symposium: Quantum Informatics for Cognitive, Social, and Semantic Processes(2010)

    2010

  59. [67]

    J. S. Bell, On the einstein podolsky rosen paradox, Physics Physique Fizika1, 195 (1964)

    1964

  60. [68]

    Uprety, Investigation and modelling of quantum-like user cognitive behaviour in information access and re- trieval (2020)

    S. Uprety, Investigation and modelling of quantum-like user cognitive behaviour in information access and re- trieval (2020)

    2020

  61. [69]

    Aerts, S

    D. Aerts, S. Aerts, J. Broekaert, and L. Gabora, The vio- lation of bell inequalities in the macroworld, Foundations of Physics30, 1387 (2000)

    2000

  62. [70]

    Aerts, J

    D. Aerts, J. A. Arguëlles, L. Beltran, S. Geriente, M. S. de Bianchi, S. Sozzo, and T. Veloz, Spin and wind direc- tions i: Identifying entanglement in nature and cognition, Foundations of Science23, 323 (2018)

    2018

  63. [71]

    Aerts, J

    D. Aerts, J. A. Arguëlles, L. Beltran, S. Geriente, M. Sas- soli de Bianchi, S. Sozzo, and T. Veloz, Spin and wind directions ii: A bell state quantum model, Foundations of Science23, 337 (2018)

    2018

  64. [72]

    Bruza, L

    P. Bruza, L. Fell, P. Hoyte, S. Dehdashti, A. Obeid, A. Gibson, and C. Moreira, Contextuality and context- sensitivity in probabilistic models of cognition, Cognitive Psychology140, 101529 (2023)

    2023

  65. [73]

    A. N. Kolmogorov, Three approaches to the quantita- tive definition of information, Problems of Information Transmission1, 1 (1965)

    1965

  66. [74]

    J. F. Clauser, M. A. Horne, A. Shimony, and R. A. Holt, Proposed experiment to test local hidden-variable theo- ries, Phys. Rev. Lett.23, 880 (1969)

    1969

  67. [75]

    Guo and A

    W. Guo and A. Caliskan, Detecting Emergent Intersec- tionalBiases: ContextualizedWordEmbeddingsContain a Distribution of Human-like Biases, inProceedings of the 2021 AAAI/ACM Conference on AI, Ethics, and Society (2021) pp. 122–133

    2021

  68. [76]

    B. K. Payne, H. A. Vuletich, and K. B. Lundberg, The Bias of Crowds: How Implicit Bias Bridges Personal and Systemic Prejudice, Psychological Inquiry28, 233 (2017)

    2017

  69. [77]

    Siddique, I

    Z. Siddique, I. Khalid, L. D. Turner, and L. Espinosa- Anke, Shifting perspectives: Steering vector ensem- bles for robust bias mitigation in llms, arXiv preprint arXiv:2503.05371 (2025)

    2025

  70. [78]

    I. O. Gallegos, R. A. Rossi, J. Barrow, M. M. Tanjim, S. Kim, F. Dernoncourt, T. Yu, R. Zhang, and N. K. Ahmed, Bias and fairness in large language models: A survey, arXiv preprint arXiv:2309.00770 (2023)

    2023 arXiv

  71. [79]

    Kitadai, K

    A. Kitadai, K. Ogawa, and N. Nishino, Examining the feasibility of large language models as survey respon- dents, in2024 IEEE International Conference on Big Data (BigData)(2024) pp. 3858–3864

    2024

  72. [80]

    Salecha, M

    A. Salecha, M. E. Ireland, S. Subrahmanya, J. Sedoc, L. H. Ungar, and J. C. Eichstaedt, Large language models display human-like social desirability biases in big five personality surveys, PNAS Nexus3, pgae533 (2024), https://academic.oup.com/pnasnexus/article- pdf/3/12/pgae53...

    2024

  73. [81]

    Tjuatja, V

    L. Tjuatja, V. Chen, T. Wu, A. Talwalkwar, and G. Neubig, Do llms exhibit human-like response bi- ases? a case study in survey design, Transactions of the Association for Computational Linguistics 12 12, 1011 (2024), https://direct.mit.edu/tacl/article- pdf/doi/10.1162/tacl_a_...

    2024 doi

  74. [82]

    S. T. Piantadosi and F. Hill, Meaning without reference in large language models, arXiv preprint arXiv:2208.02957 (2022)

    2022 arXiv

  75. [83]

    Lampinen, I

    A. Lampinen, I. Dasgupta, S. Chan, K. Mathewson, M. Tessler, A. Creswell, J. McClelland, J. Wang, and F. Hill, Can language models learn from explanations in context?, inFindings of the Association for Computa- tional Linguistics: EMNLP 2022, edited by Y. Goldberg, Z. Kozareva...

    2022

  76. [84]

    E. N. Dzhafarov, J. V. Kujala, and V. H. Cervantes, Contextuality-by-default: a brief overview of ideas, con- cepts, and terminology, inQuantum Interaction: 9th In- ternational Conference, QI 2015, Filzbach, Switzerland, July 15-17, 2015, Revised Selected Papers 9(Springer,

    2015

  77. [85]

    V. H. Cervantes and E. N. Dzhafarov, Advanced analy- sis of quantum contextuality in a psychophysical double- detection experiment, Journal of Mathematical Psychol- ogy79, 77 (2017)

    2017

  78. [86]

    J. L. Elman, Finding Structure in Time, Cognitive Sci- ence14, 179 (1990)

    1990

  79. [87]

    Wittgenstein,Philosophical Investigations(Basil Blackwell, Oxford, 1953)

    L. Wittgenstein,Philosophical Investigations(Basil Blackwell, Oxford, 1953)

    1953

  80. [88]

    W. V. O. Quine,Word & Object(MIT Press, 1960)

    1960

  81. [89]

    R. M. Hogarth and H. J. Einhorn, Order effects in belief updating: The belief-adjustment model, Cognitive Psy- chology24, 1 (1992)

    1992

  82. [90]

    J. R. Anderson and R. Milson, Human memory: An adaptive perspective, Psychological Review96, 703 (1989)

    1989

  83. [91]

    Q. Wu, G. Bansal, J. Zhang, Y. Wu, B. Li, E. Zhu, L. Jiang, X. Zhang, S. Zhang, J. Liu, A. H. Awadal- lah, R. W. White, D. Burger, and C. Wang, Au- toGen: Enabling next-gen large language model ap- plications via multi-agent conversation, arXiv preprint arXiv:2308.08155 (2023)

    2023 arXiv

  84. [92]

    LlamaIndex Contributors,Multi-Agent Workflows — LlamaIndex Documentation(2025)

    2025

  85. [93]

    Herzig, Sap’s new ai innovations and partnerships de- liver real-world results (2024)

    P. Herzig, Sap’s new ai innovations and partnerships de- liver real-world results (2024)

    2024

  86. [94]

    Srivastava, J

    J. Srivastava, J. Schuurmans, K. Ray, L. Mesnage, N. Butler, S. Dutta, T. Kubit, and T. Janner, Genai can revolutionize erp transformations, Boston Consult- ing Group Insight (2025)

    2025

  87. [95]

    Quach, A

    V. Quach, A. Fisch, T. Schuster, A. Yala, J. H. Sohn, T. S. Jaakkola, and R. Barzilay, Conformal language modeling, arXiv preprint arXiv:2306.10193 (2023)

    2023 arXiv

  88. [96]

    Z. Li, W. Wu, Y. Wang, Y. Xu, W. Hunt, and S. Stein, HMCF: A human-in-the-loop multi-robot col- laboration framework based on large language models, arXiv preprint arXiv:2505.00820 (2025)

    2025 arXiv

  89. [97]

    Naminas, Human in the loop machine learning: The key to better models (2025)

    K. Naminas, Human in the loop machine learning: The key to better models (2025)

    2025

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.