Pith. sign in

REVIEW 1 major objections 2 minor 90 references

Reviewed by Pith at T0; open to challenge.

T0 means a machine referee read the full paper against a public rubric. The mark states how deep the mechanical check went, never who wrote it. the ladder, T0–T4 →

Challenge this review Download PDF Read on arXiv ↗

T0 review · grok-4.3

Minimal physical interaction lets a robot learn to count with far less training data than vision-only systems while forming brain-like number representations.

2026-05-10 16:12 UTC

load-bearing objection Embodied motor input boosts robotic counting accuracy with little data and yields brain-like representations, but the randomization control leaves open whether training dynamics stayed truly matched across conditions. the 1 major comments →

arxiv 2604.11373 v1 submitted 2026-04-13 cs.RO cs.AI

Minimal Embodiment Enables Efficient Learning of Number Concepts in Robot

Zhegong Shangguan , Alessandro Di Nuovo , Angelo Cangelosi This is my paper
classification cs.RO cs.AI
keywords embodimentnumerical cognitionrobot learningdata efficiencycountingstructural priorneural networksdevelopmental stages
verification ladder T0 review T1 audit T2 compute T3 formal T4 reserved

The pith

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

This paper investigates how a robot can acquire abstract number concepts through basic sensorimotor experience rather than pure visual input. It shows that adding minimal embodiment to a neural network model yields much higher counting accuracy with only a small fraction of the training examples required by non-embodied baselines. The performance gain remains even after the connection between vision and movement is randomized, pointing to embodiment serving as an internal structural regularizer instead of an extra source of information. The trained model also produces number-selective units, logarithmic scaling, and rotational dynamics that match known features of biological numerical cognition. These outcomes suggest that simple bodily interaction can efficiently ground mathematical ideas in artificial agents.

Core claim

Embodied models achieve 96.8% counting accuracy with only 10% of training data, compared to 60.6% for vision-only baselines. This advantage persists when visual-motor correspondences are randomized, indicating that embodiment functions as a structural prior that regularizes learning rather than as an information source. The model spontaneously develops biologically plausible representations: number-selective units with logarithmic tuning, mental number line organization, Weber-law scaling, and rotational dynamics encoding numerical magnitude (r = 0.97, slope = 30.6 deg/count). The learning trajectory parallels children's developmental progression from subset-knowers to cardinal-principle kno

What carries the argument

A neural network trained on sequential counting via naturalistic interaction with a robotic manipulator arm, using embodiment as a structural prior that shapes the learning dynamics.

Load-bearing premise

Randomizing the visual-motor link fully isolates the regularizing effect of embodiment without introducing uncontrolled differences in network training dynamics.

What would settle it

Training both embodied and vision-only networks under identical architectures and optimization while removing all physical interaction components and checking whether the accuracy gap closes.

Watch this falsifier — get emailed when new claim-graph text bears on it.

If this is right

  • Robotic systems can acquire numerical understanding with substantially reduced data needs when minimal physical interaction is included.
  • Abstract concepts can emerge with representations that align with biological patterns such as logarithmic tuning and mental number lines.
  • The developmental sequence from subset-knowers to cardinal-principle knowers can be replicated in artificial agents through embodied training.
  • Embodiment can improve data efficiency for quantity-related tasks in human-interactive or industrial robotic applications.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • The same minimal-embodiment approach might accelerate learning of other abstract relations such as spatial or temporal ordering.
  • In safety-critical settings, quantity awareness grounded in physical interaction could reduce errors in tasks involving counting or measurement.
  • Extending the model to multi-object scenes or variable object properties would test whether the structural prior generalizes beyond simple counting.
  • Comparing the rotational dynamics observed here to those in human brain imaging during number tasks could strengthen the biological parallel.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

1 major / 2 minor

Summary. The paper claims that a neural network model trained via sensorimotor interaction on a Franka Panda robot for sequential counting achieves 96.8% accuracy with only 10% of the training data, outperforming vision-only baselines at 60.6%. This advantage persists under randomization of visual-motor correspondences, which the authors interpret as evidence that embodiment functions as a structural prior that regularizes learning rather than supplying specific information. The model spontaneously develops number-selective units exhibiting logarithmic tuning, mental number line organization, Weber-law scaling, and rotational dynamics (r=0.97, slope=30.6°/count). The learning trajectory is reported to parallel children's progression from subset-knowers to cardinal-principle knowers.

Significance. If the controls hold, the work provides empirical support for the role of minimal embodiment in grounding abstract numerical concepts with improved data efficiency and biologically aligned internal representations. The explicit randomization control is a methodological strength that allows a direct test of whether embodiment contributes beyond raw information content. Such findings could inform the design of data-efficient robotic systems for human-interactive tasks and contribute to interdisciplinary understanding of embodied cognition.

major comments (1)
  1. [Methods (randomized visual-motor condition and condition comparisons)] The central claim that embodiment acts as a structural prior rests on the performance advantage persisting in the randomized visual-motor condition. The manuscript does not provide explicit confirmation that network architecture, input dimensionality, optimization hyperparameters, data augmentation, and loss weighting remain identical across the embodied, randomized, and vision-only conditions. Retention of motor input channels (even if decorrelated) in the randomized case could alter gradient magnitudes or effective capacity relative to the vision-only baseline, which lacks motor channels entirely; this risks confounding the isolation of purely structural effects.
minor comments (2)
  1. [Abstract] The abstract reports point accuracies (96.8% and 60.6%) and a correlation (r=0.97) without accompanying variability measures, statistical tests, or references to the specific results section or figure supporting these values.
  2. [Results (representations)] Notation for the rotational dynamics (slope = 30.6°/count) would benefit from a brief definition of the angle measurement and the exact analysis used to obtain the slope.

Simulated Author's Rebuttal

1 responses · 0 unresolved

We thank the referee for their thoughtful review and for identifying the need for greater methodological clarity. We address the major comment on condition consistency below.

read point-by-point responses

  1. Referee: [Methods (randomized visual-motor condition and condition comparisons)] The central claim that embodiment acts as a structural prior rests on the performance advantage persisting in the randomized visual-motor condition. The manuscript does not provide explicit confirmation that network architecture, input dimensionality, optimization hyperparameters, data augmentation, and loss weighting remain identical across the embodied, randomized, and vision-only conditions. Retention of motor input channels (even if decorrelated) in the randomized case could alter gradient magnitudes or effective capacity relative to the vision-only baseline, which lacks motor channels entirely; this risks confounding the isolation of purely structural effects.

    Authors: We appreciate the referee highlighting this point. The embodied and randomized conditions use an identical network architecture, including the same input modules and dimensionality for visual features and motor/proprioceptive channels from the Franka Panda. Randomization affects only the pairing of visual and motor streams during training; motor channel dimensionality and structure remain unchanged. The vision-only baseline excludes motor inputs by design to isolate embodiment effects. All optimization hyperparameters, data augmentation, and loss weightings are identical across conditions. We will revise the Methods section to add an explicit paragraph confirming these equivalences and listing input dimensions per condition. While the presence of motor channels in the randomized condition does introduce an additional input stream relative to vision-only, this is intentional to test whether even decorrelated motor structure regularizes learning; the observed advantage supports our structural-prior interpretation rather than confounding it. revision: yes

Circularity Check

0 steps flagged

No significant circularity: empirical comparisons are self-contained

full rationale

The paper reports experimental results from training neural network models on a robotic counting task, comparing embodied, vision-only, and randomized visual-motor conditions. Claims of 96.8% vs 60.6% accuracy and persistence under randomization rest on direct empirical measurements and observed patterns (e.g., r=0.97 correlations, developmental trajectory parallels), not on any derivation, equation, or parameter fit that reduces to itself by construction. No self-definitional steps, fitted inputs renamed as predictions, or load-bearing self-citations appear in the provided text. The analysis is therefore self-contained against external benchmarks.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The central claim rests on standard machine-learning assumptions about neural networks learning from sensorimotor streams and on the domain assumption that embodiment supplies a structural prior; no new entities are introduced and no free parameters are explicitly fitted in the abstract.

axioms (1)
  • domain assumption Embodiment supplies a structural prior that regularizes learning of abstract concepts
    Invoked to interpret why the randomized condition still shows an advantage.

pith-pipeline@v0.9.0 · 5506 in / 1158 out tokens · 52021 ms · 2026-05-10T16:12:52.807008+00:00 · methodology

0 comments
read the original abstract

Robots are increasingly entering human-interactive scenarios that require understanding of quantity. How intelligent systems acquire abstract numerical concepts from sensorimotor experience remains a fundamental challenge in cognitive science and artificial intelligence. Here we investigate embodied numerical learning using a neural network model trained to perform sequential counting through naturalistic robotic interaction with a Franka Panda manipulator. We demonstrate that embodied models achieve 96.8\% counting accuracy with only 10\% of training data, compared to 60.6\% for vision-only baselines. This advantage persists when visual-motor correspondences are randomized, indicating that embodiment functions as a structural prior that regularizes learning rather than as an information source. The model spontaneously develops biologically plausible representations: number-selective units with logarithmic tuning, mental number line organization, Weber-law scaling, and rotational dynamics encoding numerical magnitude ($r = 0.97$, slope $= 30.6{\deg}$/count). The learning trajectory parallels children's developmental progression from subset-knowers to cardinal-principle knowers. These findings demonstrate that minimal embodiment can ground abstract concepts, improve data efficiency, and yield interpretable representations aligned with biological cognition, which may contribute to embodied mathematics tutoring and safety-critical industrial applications.

Figures

Figures reproduced from arXiv: 2604.11373 by Alessandro Di Nuovo, Angelo Cangelosi, Zhegong Shangguan.

Figure 1
Figure 1. Figure 1: Embodied numerical cognition: from human to robot. (A) Humans use pointing gestures to coordinate counting with objects. (B) Neural basis: visual (V1/V2/V3) and proprioceptive (S1/S2) signals converge in the intraparietal sulcus (IPS), supporting abstract numerical knowledge. (C) Robotic implementation: a Franka Panda manipulator performs sequential counting with a wrist-mounted camera. (D) Model architect… view at source ↗
Figure 2
Figure 2. Figure 2: Embodied learning yields superior data efficiency, distinctive learning dynamics, and structured numerical representations. a, Validation accuracy across training data fractions (10%, 50%, 100%) for embodied (blue) and vision-only (red) models, with and without ImageNet pretraining. Embodied models consistently outperform vision￾only baselines across all conditions. Notably, pretraining enhances embodied m… view at source ↗
Figure 3
Figure 3. Figure 3: Motor signals function as structural priors rather than information sources. a, Schematic of the shuffle experiment. Within each training batch, joint position labels are randomly permuted across samples while visual inputs and count labels remain unchanged. This manipulation preserves motor prediction as an auxiliary training objective but eliminates veridical visuomotor correspondence. b, Counting accura… view at source ↗
Figure 4
Figure 4. Figure 4: Curriculum learning reveals intrinsic developmental constraints in numerical acquisition. a, Schematic of three curriculum strategies. In easy-to-hard, training samples are ordered by numerosity from 1 to 10. In hard-to-easy, the order is reversed from 10 to 1. In random, samples are shuffled without systematic ordering. b, Final counting accuracy across curriculum strategies and data fractions for pretrai… view at source ↗
Figure 5
Figure 5. Figure 5: Neural population dynamics reveal number-line organization and rotational counting mechanism. (a) Tuning curves of representative number-selective units preferring numerosities 2, 5, and 8. (b) Population activity of positive (red) and negative (blue) numerosity detectors as a function of log(numerosity), showing logarithmic encoding (R 2 = 0.90 and 0.89, respectively). (c) Distribution of preferred numero… view at source ↗
Figure 6
Figure 6. Figure 6: Overview of the experimental robotic platform. A Franka Emika Panda manipulator is equipped with an Intel [PITH_FULL_IMAGE:figures/full_fig_p018_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Overview of three modeling paradigms: (a) an embodied model that integrates visual observations and motor [PITH_FULL_IMAGE:figures/full_fig_p018_7.png] view at source ↗
Figure 8
Figure 8. Figure 8: Dataset distribution. Number of counting sequences for each numerosity (1–10), following a Zipf-like distribution [PITH_FULL_IMAGE:figures/full_fig_p019_8.png] view at source ↗
Figure 9
Figure 9. Figure 9: Learning trajectory of all training configurations, presented as validation-accuracy curves over epochs in a unified [PITH_FULL_IMAGE:figures/full_fig_p020_9.png] view at source ↗
Figure 10
Figure 10. Figure 10: Heatmaps of per-number accuracy across training epochs for all experimental settings. Each panel corresponds to [PITH_FULL_IMAGE:figures/full_fig_p021_10.png] view at source ↗
Figure 11
Figure 11. Figure 11: Per-file jPCA trajectories for the 10% data condition, layer1. Each subplot corresponds to one CSV result file [PITH_FULL_IMAGE:figures/full_fig_p022_11.png] view at source ↗
Figure 12
Figure 12. Figure 12: Per-file jPCA trajectories for the 10% data condition, layer2. Each subplot corresponds to one CSV result file, [PITH_FULL_IMAGE:figures/full_fig_p023_12.png] view at source ↗
Figure 13
Figure 13. Figure 13: Per-file jPCA trajectories for the 50% data condition, layer1. Each subplot corresponds to one CSV result file, [PITH_FULL_IMAGE:figures/full_fig_p024_13.png] view at source ↗
Figure 14
Figure 14. Figure 14: Per-file jPCA trajectories for the 50% data condition, layer2. Each subplot corresponds to one CSV result file, [PITH_FULL_IMAGE:figures/full_fig_p025_14.png] view at source ↗
Figure 15
Figure 15. Figure 15: Per-file jPCA trajectories for the 100% data condition, layer1. Each subplot corresponds to one CSV result file, [PITH_FULL_IMAGE:figures/full_fig_p026_15.png] view at source ↗
Figure 16
Figure 16. Figure 16: Per-file jPCA trajectories for the 100% data condition, layer2. Each subplot corresponds to one CSV result file, [PITH_FULL_IMAGE:figures/full_fig_p027_16.png] view at source ↗

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

90 extracted references · 90 canonical work pages

  1. [1]

    M. W. Alibali and A. A. DiRusso. The function of gesture in learning to count: More than keeping track.Cognitive development, 14(1):37–56, 1999

    work page 1999

  2. [2]

    Asada.Cognitive Robotics

    M. Asada.Cognitive Robotics. Intelligent Robotics and Autonomous Agents series. MIT Press, United States, May 2022

    work page 2022

  3. [3]

    Baltruˇ saitis, C

    T. Baltruˇ saitis, C. Ahuja, and L.-P. Morency. Multi- modal machine learning: A survey and taxonomy.IEEE transactions on pattern analysis and machine intelli- gence, 41(2):423–443, 2018

    work page 2018

  4. [4]

    L. W. Barsalou. Grounded cognition.Annual review of psychology, 59:617–645, 2008

    work page 2008

  5. [5]

    Bengio, J

    Y. Bengio, J. Louradour, R. Collobert, and J. Weston. Curriculum learning. InProceedings of the 26th annual international conference on machine learning, pages 41– 48, 2009

    work page 2009

  6. [6]

    A. M. Borghi, F. Binkofski, C. Castelfranchi, F. Cimatti, C. Scorolli, and L. Tummolini. The challenge of abstract concepts.Psychological bulletin, 143(3):263, 2017

    work page 2017

  7. [7]

    S. T. Boysen. Counting in chimpanzees: Nonhuman prin- ciples and emergent properties of number. InThe devel- opment of numerical competence, pages 39–59. Psychol- ogy Press, 2014

    work page 2014

  8. [8]

    Butterworth, S

    B. Butterworth, S. Varma, and D. Laurillard. Dyscal- culia: from brain to education.science, 332(6033):1049– 1053, 2011

    work page 2011

  9. [9]

    Cangelosi.Developmental Robotics From Babies to Robots

    A. Cangelosi.Developmental Robotics From Babies to Robots. MIT Press, United States, Jan. 2015

    work page 2015

  10. [10]

    Cangelosi, A

    A. Cangelosi, A. Morse, A. Di Nuovo, M. Rucinski, F. Stramandinoli, D. Marocco, V. De La Cruz, and K. Fischer.Embodied language and number learning in developmental robots, volume Volume 2, pages 275–295. Dec. 2015

    work page 2015

  11. [11]

    S. J. Cheyette and S. T. Piantadosi. A unified account of numerosity perception.Nature human behaviour, 4(12):1265–1272, 2020

    work page 2020

  12. [12]

    M. M. Churchland, J. P. Cunningham, M. T. Kaufman, J. D. Foster, P. Nuyujukian, S. I. Ryu, and K. V. Shenoy. Neural population dynamics during reaching.Nature, 487(7405):51–56, 2012

    work page 2012

  13. [13]

    A. G. Cohn and S. M. Hazarika. Qualitative spatial rep- resentation and reasoning: An overview.Fundamenta informaticae, 46(1-2):1–29, 2001

    work page 2001

  14. [14]

    K. R. Coventry, H. B. Gudde, H. Diessel, J. Col- lier, P. Guijarro-Fuentes, M. Vulchanova, V. Vulchanov, E. Todisco, M. Reile, M. Breunesse, et al. Spatial commu- nication systems across languages reflect universal action constraints.Nature human behaviour, 7(12):2099–2110, 2023

    work page 2099

  15. [15]

    J. C. Culham and N. G. Kanwisher. Neuroimaging of cognitive functions in human parietal cortex.Current opinion in neurobiology, 11(2):157–163, 2001

    work page 2001

  16. [16]

    M. D. De Hevia, V. Izard, A. Coubart, E. S. Spelke, and A. Streri. Representations of space, time, and number in neonates.Proceedings of the National Academy of Sci- ences, 111(13):4809–4813, 2014

    work page 2014

  17. [17]

    Dehaene.The number sense: How the mind creates mathematics

    S. Dehaene.The number sense: How the mind creates mathematics. OUP USA, 2011

    work page 2011

  18. [18]

    Dehaene, M

    S. Dehaene, M. Piazza, P. Pinel, and L. Cohen. Three parietal circuits for number processing. InThe hand- book of mathematical cognition, pages 433–453. Psychol- ogy Press, 2005

    work page 2005

  19. [19]

    K. Devlin. The mathematical brain.Mind, brain, and education, 2010

    work page 2010

  20. [20]

    Di Nuovo and J

    A. Di Nuovo and J. L. McClelland. Developing the knowl- edge of number digits in a child-like robot.Nature Ma- chine Intelligence, 1:594–605, 2019

    work page 2019

  21. [21]

    Durstewitz, G

    D. Durstewitz, G. Koppe, and M. I. Thurm. Reconstruct- ing computational system dynamics from neural data with recurrent neural networks.Nature Reviews Neu- roscience, 24(11):693–710, 2023

    work page 2023

  22. [22]

    J. L. Elman. Finding structure in time.Cognitive Sci- ence, 14(2):179–211, 1990

    work page 1990

  23. [23]

    J. L. Elman. Learning and development in neural net- works: The importance of starting small.Cognition, 48(1):71–99, 1993

    work page 1993

  24. [24]

    J. Elsner. Taming the panda with python: A powerful duo for seamless robotics programming and integration. SoftwareX, 24:101532, 2023

    work page 2023

  25. [25]

    M. Fang, Z. Zhou, S. Y. Chen, and J. L. McClelland. Can a recurrent neural network learn to count things? InProceedings of the Annual Meeting of the Cognitive Science Society, volume 40. Cognitive Science Society, 2018

    work page 2018

  26. [26]

    Feigenson, S

    L. Feigenson, S. Dehaene, and E. Spelke. Core systems of number.Trends in cognitive sciences, 8(7):307–314, 2004

    work page 2004

  27. [27]

    D. J. Felleman and D. C. Van Essen. Distributed hierar- chical processing in the primate cerebral cortex.Cerebral cortex (New York, NY: 1991), 1(1):1–47, 1991

    work page 1991

  28. [28]

    M. H. Fischer and S. Shaki. Spatial associations in numerical cognition—from single digits to arith- metic.Quarterly journal of experimental psychology, 67(8):1461–1483, 2014

    work page 2014

  29. [29]

    M. H. Fischer and R. A. Zwaan. Embodied language: A review of the role of the motor system in language comprehension.The Quarterly Journal of Experimental Psychology, 61(6):825–850, 2008

    work page 2008

  30. [30]

    Gelman and C

    R. Gelman and C. Gallistel.The Child’s Understanding of Number. Harvard University Press, 1986

    work page 1986

  31. [31]

    Gennari, S

    G. Gennari, S. Dehaene, C. Valera, and G. Dehaene- Lambertz. Spontaneous supra-modal encoding of number in the infant brain.Current Biology, 33(10):1906–1915, 2023

    work page 1906

  32. [32]

    Gertner, C

    Y. Gertner, C. Fisher, and J. Eisengart. Learning words and rules: Abstract knowledge of word order in early sen- tence comprehension.Psychological science, 17(8):684– 691, 2006

    work page 2006

  33. [33]

    Grefkes and G

    C. Grefkes and G. R. Fink. The functional organization of the intraparietal sulcus in humans and monkeys.Journal of anatomy, 207(1):3–17, 2005

    work page 2005

  34. [34]

    B. M. Harvey, B. P. Klein, N. Petridou, and S. O. Du- moulin. Topographic representation of numerosity in the human parietal cortex.Science, 341(6150):1123–1126, 2013

    work page 2013

  35. [35]

    Hochreiter and J

    S. Hochreiter and J. Schmidhuber. Long short-term memory.Neural computation, 9(8):1735–1780, 1997

    work page 1997

  36. [36]

    J. J. Hopfield. Neural networks and physical systems with emergent collective computational abilities.Pro- ceedings of the national academy of sciences, 79(8):2554– 2558, 1982. 15

    work page 1982

  37. [37]

    Høyrup.In measure, number, and weight: Studies in mathematics and culture

    J. Høyrup.In measure, number, and weight: Studies in mathematics and culture. SUNY Press, 1994

    work page 1994

  38. [38]

    E. M. Hubbard, M. Piazza, P. Pinel, and S. Dehaene. In- teractions between number and space in parietal cortex. Nature reviews neuroscience, 6(6):435–448, 2005

    work page 2005

  39. [39]

    D. H. Hubel and T. N. Wiesel. Receptive fields and func- tional architecture of monkey striate cortex.The Journal of physiology, 195(1):215–243, 1968

    work page 1968

  40. [40]

    Izard, C

    V. Izard, C. Sann, E. S. Spelke, and A. Streri. New- born infants perceive abstract numbers.Proceedings of the National Academy of Sciences, 106(25):10382–10385, 2009

    work page 2009

  41. [41]

    Jitsuishi and A

    T. Jitsuishi and A. Yamaguchi. Identification of a distinct association fiber tract “ips-fg” to connect the intrapari- etal sulcus areas and fusiform gyrus by white matter dis- section and tractography.Scientific reports, 10(1):15475, 2020

    work page 2020

  42. [42]

    E. L. Kaufman, M. W. Lord, T. W. Reese, and J. Volk- mann. The discrimination of visual number.The Amer- ican journal of psychology, 62(4):498–525, 1949

    work page 1949

  43. [43]

    G. Kim, J. Jang, S. Baek, M. Song, and S.-B. Paik. Visual number sense in untrained deep neural networks.Science advances, 7(1):eabd6127, 2021

    work page 2021

  44. [44]

    Kosmas, A

    P. Kosmas, A. Ioannou, and P. Zaphiris. Implementing embodied learning in the classroom: Effects on children’s memory and language skills.Educational Media Interna- tional, 56(1):59–74, 2019

    work page 2019

  45. [45]

    Kriegeskorte, M

    N. Kriegeskorte, M. Mur, and P. A. Bandettini. Repre- sentational similarity analysis-connecting the branches of systems neuroscience.Frontiers in systems neuroscience, 2:249, 2008

    work page 2008

  46. [46]

    Krizhevsky, I

    A. Krizhevsky, I. Sutskever, and G. E. Hinton. Imagenet classification with deep convolutional neural networks. Communications of the ACM, 60(6):84–90, 2012

    work page 2012

  47. [47]

    Kulkarni, V

    G. Kulkarni, V. Premraj, V. Ordonez, S. Dhar, S. Li, Y. Choi, A. C. Berg, and T. L. Berg. Babytalk: Un- derstanding and generating simple image descriptions. IEEE transactions on pattern analysis and machine in- telligence, 35(12):2891–2903, 2013

    work page 2013

  48. [48]

    Lahat, T

    D. Lahat, T. Adali, and C. Jutten. Multimodal data fu- sion: an overview of methods, challenges, and prospects. Proceedings of the IEEE, 103(9):1449–1477, 2015

    work page 2015

  49. [49]

    Lakoff and M

    G. Lakoff and M. Johnson.Metaphors we live by. Uni- versity of Chicago press, 1980

    work page 1980

  50. [50]

    Le Corre and S

    M. Le Corre and S. Carey. One, two, three, four, nothing more: An investigation of the conceptual sources of the verbal counting principles.Cognition, 105(2):395–438, 2007

    work page 2007

  51. [51]

    Le Corre, P

    M. Le Corre, P. Li, B. H. Huang, G. Jia, and S. Carey. Numerical morphology supports early number word learning: Evidence from a comparison of young mandarin and english learners.Cognitive psychology, 88:162–186, 2016

    work page 2016

  52. [52]

    LeCun, Y

    Y. LeCun, Y. Bengio, and G. Hinton. Deep learning. nature, 521(7553):436–444, 2015

    work page 2015

  53. [53]

    Y. Li, Q. Gao, T. Zhao, B. Wang, H. Sun, H. Lyu, R. D. Hawkins, N. Vasconcelos, T. Golan, D. Luo, and H. Deng. Core knowledge deficits in multi-modal language mod- els. In A. Singh, M. Fazel, D. Hsu, S. Lacoste-Julien, F. Berkenkamp, T. Maharaj, K. Wagstaff, and J. Zhu, editors,Proceedings of the 42nd International Confer- ence on Machine Learning, volume...

    work page 2025

  54. [54]

    D. A. Liao, K. F. Brecht, L. Veit, and A. Nieder. Crows “count” the number of self-generated vocalizations.Sci- ence, 384(6698):874–877, 2024

    work page 2024

  55. [55]

    Lindemann, A

    O. Lindemann, A. Alipour, and M. H. Fischer. Finger counting habits modulate spatial-numerical associations. Cortex, 43(3):386–392, 2007

    work page 2007

  56. [56]

    Loetscher, U

    T. Loetscher, U. Schwarz, M. Schubiger, and P. Brugger. Head turns bias random number generation.Current Bi- ology, 20(14):R619–R620, 2010

    work page 2010

  57. [57]

    Malafouris

    L. Malafouris. Grasping the concept of number: how did the sapient mind move beyond approximation.The ar- chaeology of measurement: Comprehending heaven, earth and time in ancient societies, pages 35–42, 2010

    work page 2010

  58. [58]

    A. Nieder. Prefrontal cortex and the evolution of symbolic reference.Current Opinion in Neurobiology, 19(1):99–108, 2009

    work page 2009

  59. [59]

    A. Nieder. The calculating brain.Physiological reviews, 105(1):267–314, 2025

    work page 2025

  60. [60]

    Nieder and E

    A. Nieder and E. K. Miller. Representation of number in the brain.Annual review of neuroscience, 29:561–581, 2006

    work page 2006

  61. [61]

    OSF repository:https://osf.io/jk4u8/overview? view_only=95fcde69554045788995b8ab2fdabc0d

    work page

  62. [62]

    Pecyna and A

    L. Pecyna and A. Cangelosi. Influence of pointing on learning to count: A neuro-robotics model. In2018 IEEE Symposium Series on Computational Intelligence (SSCI), pages 358–365, 2018

    work page 2018

  63. [63]

    Pecyna, A

    L. Pecyna, A. Cangelosi, and A. Di Nuovo. A robot that counts like a child: A developmental model of counting and pointing.Psychological Research, Sept. 2020

    work page 2020

  64. [64]

    Pellegrino and A

    A. Pellegrino and A. Chadwick. Rnns perform task computations by dynamically warping neural representa- tions. In39th Conference on Neural Information Process- ing Systems, Proceedings. Curran Associates Inc, Sept. 2025

    work page 2025

  65. [65]

    Piaget.The child’s conception of number.WW Norton & Co., 1965

    J. Piaget.The child’s conception of number.WW Norton & Co., 1965

    work page 1965

  66. [66]

    Piaget and B

    J. Piaget and B. Inhelder.The child’s conception of space. Norton, 1967

    work page 1967

  67. [67]

    Proverbio and M

    A. Proverbio and M. Carminati. Finger-counting obser- vation interferes with number processing.Neuropsycholo- gia, 131:275–284, 2019

    work page 2019

  68. [68]

    Ranzini, S

    M. Ranzini, S. Brigadoi, Y. Din¸ c, P. Scatturin, A. Vallesi, C. Semenza, U. Castiello, M. Zorzi, and S. Cutini. Un- veiling the sensorimotor basis of numerical processing: A functional near-infrared spectroscopy (fnirs) study.Neu- roImage, page 121576, 2025

    work page 2025

  69. [69]

    S. K. Revkin, M. Piazza, V. Izard, L. Cohen, and S. De- haene. Does subitizing reflect numerical estimation? Psychological science, 19(6):607–614, 2008

    work page 2008

  70. [70]

    Rizzolatti and L

    G. Rizzolatti and L. Craighero. The mirror-neuron sys- tem.Annu. Rev. Neurosci., 27(1):169–192, 2004

    work page 2004

  71. [71]

    Rodriguez and J

    P. Rodriguez and J. Wiles. Recurrent neural networks can learn to implement symbol-sensitive counting.Ad- vances in Neural Information Processing Systems, 10, 1997

    work page 1997

  72. [72]

    Roesch, K

    S. Roesch, K. Moeller, and J. Bahnmueller. Finger count- ing, finger number gesturing, and basic numerical skills: A cross-sectional study in 3-to 5-year-olds.Journal of experimental child psychology, 242:105892, 2024

    work page 2024

  73. [73]

    Ruci´ nski, A

    M. Ruci´ nski, A. Cangelosi, and T. Belpaeme. Robotic model of the contribution of gesture to learning to count. In2012 IEEE International Conference on Development and Learning and Epigenetic Robotics (ICDL), pages 1– 16 6, 2012

    work page 2012

  74. [74]

    D. A. Sabatini and M. T. Kaufman. Reach-dependent reorientation of rotational dynamics in motor cortex.Na- ture Communications, 15(1):7007, 2024

    work page 2024

  75. [75]

    B. W. Sarnecka and M. D. Lee. Levels of number knowl- edge during early childhood.Journal of experimental child psychology, 103(3):325–337, 2009

    work page 2009

  76. [76]

    R. R. Selvaraju, M. Cogswell, A. Das, R. Vedantam, D. Parikh, and D. Batra. Grad-cam: Visual explana- tions from deep networks via gradient-based localization. InProceedings of the IEEE international conference on computer vision, pages 618–626, 2017

    work page 2017

  77. [77]

    K. V. Shenoy, M. Sahani, and M. M. Churchland. Cor- tical control of arm movements: a dynamical systems perspective.Annual review of neuroscience, 36:337–359, 2013

    work page 2013

  78. [78]

    Spelke.What babies know: Core knowledge and com- position volume 1, volume 1

    E. Spelke.What babies know: Core knowledge and com- position volume 1, volume 1. Oxford University Press, 2022

    work page 2022

  79. [79]

    Stoianov and M

    I. Stoianov and M. Zorzi. Emergence of a’visual number sense’in hierarchical generative models.Nature neuro- science, 15(2):194–196, 2012

    work page 2012

  80. [80]

    J. A. Thompson, H. Sheahan, T. Dumbalska, J. D. Sand- brink, M. Piazza, and C. Summerfield. Zero-shot count- ing with a dual-stream neural network model.Neuron, 112(24):4147–4158.e5, 2024

    work page 2024

Showing first 80 references.