Pith. sign in

REVIEW 2 major objections 1 minor 55 references

Bidirectional Tutoring for Developmental Motor Learning in Robots: Co-Developed Interaction Dynamics Support Stable Learning

T0 review · 2 major / 1 minor · reviewed 2026-06-26 · grok-4.3

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

Pith's one-line read Bidirectional tutoring enables robots to develop consistent motor behaviors via co-adapted interaction dynamics.

desk verdict The abstract lays out a clear bidirectional tutoring hypothesis for robot motor learning but gives zero quantitative results or analysis to check if the experiments actually support it. read the letter →

arxiv 2606.19728 v1 pith:36XJJW6R submitted 2026-06-18 cs.RO cs.AI

Rui Fukushima , Jun Tani This is my paper · ORCID
classification cs.ROcs.AI
keywords bidirectionaltutoringdevelopmentalmotorlearninghuman-robotinteractionfreeenergyprinciplegenerativereplayobjectmanipulationconsistentbehaviors
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

The paper tests whether bidirectional tutoring between a robot and tutor produces more stable motor learning than standard unidirectional demonstrations. It argues that the robot's prior experiences serve as constraints on jointly developed trajectories, yielding coherent patterns and better generalization. Experiments with a humanoid robot on an object manipulation task, using both human tutors and a controlled AI tutor, show consistent behaviors emerging stage by stage along with a gradual drop in required guidance. This setup mirrors infant-caregiver dynamics and employs a free-energy-principle neural network with generative replay to support learning from single episodes.

What carries the argument

Bidirectional tutoring, the process in which tutor and robot mutually adapt so that the robot's accumulated experiences constrain the shared behavioral trajectories.

What would settle it

If bidirectional tutoring experiments produced the same behavioral inconsistency and sustained tutor dependence seen in unidirectional setups, the central claim would be falsified.

Watch

Extended reading notes

Core claim

The paper claims that bidirectional tutoring fosters consistent behaviors and stage-wise generalization in a robot's object manipulation task by allowing the robot's past experiences to act as prior constraints on co-developed interaction trajectories, implemented via a free-energy-principle-based neural network with generative replay that enables stable learning from single episodes, and this effect holds in both human-robot and AI-robot settings where the robot gradually requires less tutor guidance.

Load-bearing premise

The robot's past experiences function as prior constraints that shape the dynamics of their co-developed trajectories in bidirectional interaction.

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, simulated authors' rebuttal, and a circularity audit.

Referee Report

2 major / 1 minor

Summary. The paper hypothesizes that bidirectional tutoring in robot motor learning supplies prior constraints from the robot's past experiences (via an FEP-based neural network with generative replay) that promote consistent behavioral patterns and stage-wise generalization, unlike unidirectional interaction. It tests this via two physical-robot experiments on an object-manipulation task—one with a human tutor and one with an AI tutor using an adaptive intervention mechanism—reporting that bidirectional conditions produced consistent behaviors, generalization across stages, and decreasing tutor guidance.

Significance. If the experimental claims hold with adequate quantitative support, the work would strengthen the case for socially grounded, bidirectional frameworks in developmental robotics and demonstrate a practical implementation of FEP plus generative replay for stable, single-episode sequence learning. The dual human/AI-tutor design is a positive feature for testing robustness under controlled conditions.

major comments (2)
  1. [Results/Discussion] Results and Discussion sections: the reported outcomes (consistent behaviors, stage-wise generalization, reduced tutor guidance) are described qualitatively at a high level with no quantitative metrics, error bars, statistical tests, or direct comparisons to unidirectional baselines. This prevents assessment of whether the data actually support the central hypothesis.
  2. [Methods] Methods, AI-tutor experiment: the adaptive intervention mechanism and how it operationalizes 'bidirectional' vs. 'unidirectional' conditions are not specified in sufficient detail (e.g., exact intervention rules, state representations, or replay buffer mechanics) to allow replication or to confirm that past experiences function as prior constraints as claimed in the abstract.
minor comments (1)
  1. [Abstract/Introduction] Abstract and introduction: the phrasing 'the robot gradually required less tutor guidance' should be accompanied by a precise operational definition of 'guidance' and how it was measured.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for the constructive feedback, which identifies key areas where additional rigor and detail will strengthen the manuscript. We respond point-by-point to the major comments below.

read point-by-point responses

  1. Referee: [Results/Discussion] Results and Discussion sections: the reported outcomes (consistent behaviors, stage-wise generalization, reduced tutor guidance) are described qualitatively at a high level with no quantitative metrics, error bars, statistical tests, or direct comparisons to unidirectional baselines. This prevents assessment of whether the data actually support the central hypothesis.

    Authors: We agree that the current Results and Discussion rely on qualitative descriptions without supporting quantitative analysis. In the revised version we will extract and report quantitative metrics from the existing trial data, including behavioral consistency (e.g., standard deviation of key trajectory features across episodes), stage-wise generalization success rates, and tutor-intervention counts per stage, each accompanied by error bars and appropriate statistical tests. Direct unidirectional baselines were not collected in the reported experiments; we will therefore add an explicit limitations paragraph explaining the design choice and its implications rather than fabricating comparisons. revision: partial

  2. Referee: [Methods] Methods, AI-tutor experiment: the adaptive intervention mechanism and how it operationalizes 'bidirectional' vs. 'unidirectional' conditions are not specified in sufficient detail (e.g., exact intervention rules, state representations, or replay buffer mechanics) to allow replication or to confirm that past experiences function as prior constraints as claimed in the abstract.

    Authors: We accept that the Methods section for the AI-tutor experiment currently lacks the granularity needed for replication. The revised manuscript will expand this section with the precise intervention rules (prediction-error thresholds and timing), the state representation fed to the tutor, and the generative-replay buffer update mechanics, thereby clarifying how bidirectional dynamics are realized and how past experiences operate as priors. revision: yes

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity; empirical results from robot trials

full rationale

The paper advances a testable hypothesis about bidirectional tutoring via two physical robot experiments (human and AI tutor) and reports observed outcomes on consistency, generalization, and reduced guidance needs. The framework uses an FEP-based network with generative replay, but the central claims rest on experimental data rather than any derivation, equation, or fitted parameter that reduces to its own inputs by construction. No self-citation chain or ansatz is shown to be load-bearing for the reported results.

Assumptions & free parameters 0 free parameters · 1 assumptions · 0 invented entities

Abstract provides insufficient detail for exhaustive ledger; the primary domain assumption is the applicability of the FEP-based model to developmental motor learning.

assumptions (1)
  • domain assumption Free-energy-principle-based neural network extended with generative replay supports stable sequence-by-sequence learning from single tutored episodes.
    Invoked as the core implementation framework in the abstract without additional justification or derivation.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Bidirectional Tutoring for Developmental Motor Learning in Robots: Co-Developed Interaction Dynamics Support Stable Learning." pith.science (2026). https://pith.science/paper/36XJJW6R

@misc{pith2026260619728,
  author       = {Pith},
  title        = {Pith review of: Bidirectional Tutoring for Developmental Motor Learning in Robots: Co-Developed Interaction Dynamics Support Stable Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/36XJJW6R}},
  note         = {Machine review of arXiv:2606.19728}
}
read the original abstract

Infants are well known to develop their motor skills through dense interaction with caregivers. Although such social interaction is crucial for human development, motor-skill learning in robots is often treated as a unidirectional process in which robots passively receive demonstrations from tutors. This overlooks a key property of social interaction: it is inherently bidirectional, with tutor and learner dynamically adapting to each other. In such interactions, the robot's past experiences may function as prior constraints that shape the dynamics of their co-developed trajectories. We hypothesize that bidirectional tutoring allows such constraints to guide the formation of consistent behavioral patterns that preserve behavioral coherence and support generalization, whereas unidirectional interaction lacks such constraints and leads to broader, less consistent behavioral patterns. To examine this hypothesis, we conducted two experiments with a physical humanoid robot performing an object manipulation task: one involving human-robot interaction and another employing an AI tutor interacting with the real robot through an adaptive intervention mechanism designed to examine whether similar effects would emerge under more controlled conditions. We implement the developmental learning framework using a free-energy-principle-based neural network extended with generative replay, which supports stable sequence-by-sequence learning from single tutored episodes. Across both settings, bidirectional tutoring fostered consistent behaviors and stage-wise generalization, while the robot gradually required less tutor guidance. These results suggest that bidirectional tutoring, as an embodied and socially grounded approach, provides an effective scaffold for developmental motor learning in robots.

Figures

Figures reproduced from arXiv: 2606.19728 by the authors.

Figure 1
Figure 1. Conceptual diagram of unidirectional and bidirectional tutoring in developmental motor learning. The upper interaction schematics illustrate how tutoring trajectories are generated in each condition, and the middle panels overlay the resulting trajectories across developmental phases. The speech bubbles contrast unidirectional tutoring, where the tutor demonstrates the task while the robot passively follows, with bi… view at source ↗
Figure 2
Figure 2. Overview of experimental setups and task settings a Torobo humanoid robot. b Example of a successful task sequence: 1) A robot begins in a home posture with a red cylindrical object placed at the center of the workspace indicated by a black circle; 2) the experimenter moves the red object to a random position; 3) the robot begins to reach for and grasp the object; 4) the robot lifts up the object; 5) the robot place… view at source ↗
Figure 3
Figure 3. Overview of tutoring setups a Diagram of unidirectional tutoring, where the human or AI tutor provides one-way guidance to the robot. b Diagram of bidirectional tutoring, where the robot generates actions based on its own predictions while the tutor intervenes as needed. c Example of a human-robot tutoring sequence. 1-2) The object is moved to a random position; 3-5) the tutor provides physical guidance for the robo… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Illustration of the computational flow of the PV-RNN during training and online inference. a Training process. The model generates predictions over time through recurrent computation. At each time step, the reconstruction error et is computed from the prediction x¯t an…
Figure 5
Figure 5. Figure 5: Success rate across developmental phases for the unidirectional (Uni) and bidirectional (Bi) tutoring conditions. Data points represent the mean success rate across training sets A–C, and error bars indicate the standard error of the mean (SEM). Two spatial error thres…
Figure 6
Figure 6. Figure 6: Qualitative and quantitative comparison of tutoring dynamics under the unidirectional and bidirectional conditions. a Left end-effector position and rotation trajectories during tutoring, overlaid across all tutoring phases for each training set (A–C). Each trace corre…
Figure 7
Figure 7. Figure 7: Changes in human tutor intervention during bidirectional tutoring. Box plots show the total excess torque applied by the human tutor during each tutoring episode in the bidirectional condition. Data from learning sets A–C were pooled within each developmental phase. Th…
Figure 8
Figure 8. Figure 8: Example of variability in AI-tutor-generated trajectories under the same object-location condition. Panels (1)–(4) show overlaid snapshots from three repeated executions at representative time points: (1) reaching and grasping at t=200, (2) lift-up at t=400, (3) put-do…
Figure 9
Figure 9. Figure 9: Schematic illustration of AI-tutor intervention through online inference in the bidirectional condition. (a) Low expected movement variability case, where ν is small in the AI tutor model. (b) High expected movement variability case, where ν is large in the AI tutor mo…
Figure 10
Figure 10. Figure 10: Success rate across developmental phases in the AI tutoring experiment. Success rates are shown for the unidirectional (Uni) and bidi￾rectional (Bi) tutoring conditions. Data points represent the mean success rate across training sets A-C, and error bars indicate the …
Figure 12
Figure 12. Figure 12: Representative examples of AI-tutoring dynamics at Phase 10. a Unidirectional conditions. b Bidirectional conditions. The examples were selected from the same object-position set. Snapshots are shown at two representative time points: during reaching and grasping (t=2…
Figure 13
Figure 13. Figure 13: Qualitative and quantitative comparison of AI-tutoring dynamics under the unidirectional and bidirectional conditions. a Left end-effector position and rotation trajectories obtained during AI-tutoring sessions, overlaid across the ten developmental phases for each tr…
Figure 14
Figure 14. Figure 14: Changes in AI tutor intervention during bidirectional tutoring. Box plots show the total intervention weight applied as tutor intervention during each tutoring episode in the bidirectional condition. Data from training sets A–C were pooled within each developmental ph…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

55 extracted references · 4 canonical work pages

  1. [1]

    Imitation of facial and manual gestures by human neonates,

    A. N. Meltzoff and M. K. Moore, “Imitation of facial and manual gestures by human neonates,”Science, vol. 198, no. 4312, pp. 75–78, 1977

    1977

  2. [2]

    J. M. PINE, “Tomasello, m., constructing a language: a usage-based theory of language acquisition. cambridge, ma: Harvard university press,

  3. [3]

    pp. 388. hardback,£ 29.95. isbn 0-674-01030-2.”Journal of Child Language, vol. 32, no. 3, pp. 697–702, 2005. 15

    2005

  4. [4]

    Participatory sense-making: An enactive approach to social cognition,

    H. De Jaegher and E. Di Paolo, “Participatory sense-making: An enactive approach to social cognition,”Phenomenology and the cognitive sciences, vol. 6, no. 4, pp. 485–507, 2007

    2007

  5. [5]

    The role of tutoring in problem solving,

    D. Wood, J. S. Bruner, and G. Ross, “The role of tutoring in problem solving,”Journal of child psychology and psychiatry, vol. 17, no. 2, pp. 89–100, 1976

    1976

  6. [6]

    A. N. Meltzoff and W. Prinz,The imitative mind: Development, evolution and brain bases. Cambridge University Press, 2002, vol. 6

    2002

  7. [7]

    Social situatedness of natural and artificial intelligence: Vygotsky and beyond,

    J. Lindblom and T. Ziemke, “Social situatedness of natural and artificial intelligence: Vygotsky and beyond,”Adaptive Behavior, vol. 11, no. 2, pp. 79–96, 2003

    2003

  8. [8]

    L. S. Vygotsky,Mind in society: The development of higher psycholog- ical processes. Harvard university press, 1978, vol. 86

    1978

Show all 55 references
  1. [9]

    L. S. Vygotsky,Thought and language. MIT press, 2012, vol. 29

    2012

  2. [10]

    Developmental robotics: a survey,

    M. Lungarella, G. Metta, R. Pfeifer, and G. Sandini, “Developmental robotics: a survey,”Connection science, vol. 15, no. 4, pp. 151–190, 2003

    2003

  3. [11]

    Cognitive developmental robotics: A survey,

    M. Asada, K. Hosoda, Y . Kuniyoshi, H. Ishiguro, T. Inui, Y . Yoshikawa, M. Ogino, and C. Yoshida, “Cognitive developmental robotics: A survey,”IEEE transactions on autonomous mental development, vol. 1, no. 1, pp. 12–34, 2009

    2009

  4. [12]

    From babies to robots: the contri- bution of developmental robotics to developmental psychology,

    A. Cangelosi and M. Schlesinger, “From babies to robots: the contri- bution of developmental robotics to developmental psychology,”Child Development Perspectives, vol. 12, no. 3, pp. 183–188, 2018

    2018

  5. [13]

    Staged development of robot skills: Behavior formation, affordance learning and imitation with motionese,

    E. Ugur, Y . Nagai, E. Sahin, and E. Oztop, “Staged development of robot skills: Behavior formation, affordance learning and imitation with motionese,”IEEE Transactions on Autonomous Mental Development, vol. 7, no. 2, pp. 119–139, 2015

    2015

  6. [14]

    Learning from demonstration,

    S. Schaal, “Learning from demonstration,”Advances in neural informa- tion processing systems, vol. 9, 1996

    1996

  7. [15]

    A survey of robot learning from demonstration,

    B. D. Argall, S. Chernova, M. Veloso, and B. Browning, “A survey of robot learning from demonstration,”Robotics and autonomous systems, vol. 57, no. 5, pp. 469–483, 2009

    2009

  8. [16]

    Is imitation learning the route to humanoid robots?

    S. Schaal, “Is imitation learning the route to humanoid robots?”Trends in cognitive sciences, vol. 3, no. 6, pp. 233–242, 1999

    1999

  9. [17]

    Physical human–robot interaction,

    S. Haddadin and E. Croft, “Physical human–robot interaction,” in Springer handbook of robotics. Springer, 2016, pp. 1835–1874

    2016

  10. [18]

    Learning to reproduce fluctuating time series by inferring their time-dependent stochastic properties: Application in robot learning via tutoring,

    S. Murata, J. Namikawa, H. Arie, S. Sugano, and J. Tani, “Learning to reproduce fluctuating time series by inferring their time-dependent stochastic properties: Application in robot learning via tutoring,”IEEE Transactions on Autonomous Mental Development, vol. 5, no. 4, pp. 2...

    2013

  11. [19]

    Learning and comfort in human– robot interaction: A review,

    W. Wang, Y . Chen, R. Li, and Y . Jia, “Learning and comfort in human– robot interaction: A review,”Applied Sciences, vol. 9, no. 23, p. 5152, 2019

    2019

  12. [20]

    Discriminative and adaptive imitation in uni-manual and bi-manual tasks,

    A. G. Billard, S. Calinon, and F. Guenter, “Discriminative and adaptive imitation in uni-manual and bi-manual tasks,”Robotics and Autonomous Systems, vol. 54, no. 5, pp. 370–384, 2006

    2006

  13. [21]

    Imitation learning of positional and force skills demonstrated via kinesthetic teaching and haptic input,

    P. Kormushev, S. Calinon, and D. G. Caldwell, “Imitation learning of positional and force skills demonstrated via kinesthetic teaching and haptic input,”Advanced Robotics, vol. 25, no. 5, pp. 581–603, 2011

    2011

  14. [22]

    Dynamic and interactive generation of object handling behaviors by a small humanoid robot using a dynamic neural network model,

    M. Ito, K. Noda, Y . Hoshino, and J. Tani, “Dynamic and interactive generation of object handling behaviors by a small humanoid robot using a dynamic neural network model,”Neural Networks, vol. 19, no. 3, pp. 323–337, 2006

    2006

  15. [23]

    Self-organization of behavioral primitives as multiple attractor dynamics: A robot experiment,

    J. Tani and M. Ito, “Self-organization of behavioral primitives as multiple attractor dynamics: A robot experiment,”IEEE Transactions on Systems, Man, and Cybernetics-Part A: Systems and Humans, vol. 33, no. 4, pp. 481–488, 2003

    2003

  16. [24]

    Generalization capability for imitation learning,

    Y . Wang, “Generalization capability for imitation learning,”arXiv preprint arXiv:2504.18538, 2025

    2025

  17. [25]

    Consistency matters: Defining demonstration data quality metrics in robot learning from demonstration,

    M. Sakr, J. Zhang, H. M. V . d. Loos, D. Kuli ´c, and E. Croft, “Consistency matters: Defining demonstration data quality metrics in robot learning from demonstration,”ACM Transactions on Human-Robot Interaction, vol. 15, no. 2, pp. 1–31, 2025

    2025

  18. [26]

    Behavior transformers: Cloningkmodes with one stone,

    N. M. Shafiullah, Z. Cui, A. A. Altanzaya, and L. Pinto, “Behavior transformers: Cloningkmodes with one stone,”Advances in neural information processing systems, vol. 35, pp. 22 955–22 968, 2022

    2022

  19. [27]

    The bliss (not the problem) of motor abundance (not redundancy),

    M. L. Latash, “The bliss (not the problem) of motor abundance (not redundancy),”Experimental brain research, vol. 217, no. 1, pp. 1–5, 2012

    2012

  20. [28]

    The role of execution noise in movement variability,

    R. J. Van Beers, P. Haggard, and D. M. Wolpert, “The role of execution noise in movement variability,”Journal of neurophysiology, vol. 91, no. 2, pp. 1050–1063, 2004

    2004

  21. [29]

    Temporal structure of motor variability is dynamically regulated and predicts motor learning ability,

    H. G. Wu, Y . R. Miyamoto, L. N. G. Castro, B. P. ¨Olveczky, and M. A. Smith, “Temporal structure of motor variability is dynamically regulated and predicts motor learning ability,”Nature neuroscience, vol. 17, no. 2, pp. 312–321, 2014

    2014

  22. [30]

    A theory of cortical responses,

    K. Friston, “A theory of cortical responses,”Philosophical transactions of the Royal Society B: Biological sciences, vol. 360, no. 1456, pp. 815–836, 2005

    2005

  23. [31]

    Clark,Surfing uncertainty: Prediction, action, and the embodied mind

    A. Clark,Surfing uncertainty: Prediction, action, and the embodied mind. Oxford University Press, 2015

    2015

  24. [32]

    Hohwy,The predictive mind

    J. Hohwy,The predictive mind. OUP Oxford, 2013

    2013

  25. [33]

    A novel predictive-coding-inspired variational rnn model for online prediction and recognition,

    A. Ahmadi and J. Tani, “A novel predictive-coding-inspired variational rnn model for online prediction and recognition,”Neural computation, vol. 31, no. 11, pp. 2025–2074, 2019

    2025

  26. [34]

    Deep active inference in physical human-robot interaction: Balancing explo- ration and goal-directed behavior,

    J. Borojevic, G. W. Haddon-Hill, J. Sandoval, and S. Murata, “Deep active inference in physical human-robot interaction: Balancing explo- ration and goal-directed behavior,” in2026 IEEE/SICE International Symposium on System Integration (SII). IEEE, 2026, pp. 1504–1509

    2026

  27. [35]

    Codevelopmental learning between human and humanoid robot using a dynamic neural- network model,

    J. Tani, R. Nishimoto, J. Namikawa, and M. Ito, “Codevelopmental learning between human and humanoid robot using a dynamic neural- network model,”IEEE Transactions on Systems, Man, and Cybernetics, Part B (Cybernetics), vol. 38, no. 1, pp. 43–59, 2008

    2008

  28. [36]

    Trajectories and keyframes for kinesthetic teaching: A human-robot interaction per- spective,

    B. Akgun, M. Cakmak, J. W. Yoo, and A. L. Thomaz, “Trajectories and keyframes for kinesthetic teaching: A human-robot interaction per- spective,” inProceedings of the seventh annual ACM/IEEE international conference on Human-Robot Interaction, 2012, pp. 391–398

    2012

  29. [37]

    Physical human-robot interaction: Mutual learning and adaptation,

    S. Ikemoto, H. B. Amor, T. Minato, B. Jung, and H. Ishiguro, “Physical human-robot interaction: Mutual learning and adaptation,”IEEE robotics & automation magazine, vol. 19, no. 4, pp. 24–35, 2012

    2012

  30. [38]

    Incremental learning of goal- directed actions in a dynamic environment by a robot using active inference,

    T. Matsumoto, W. Ohata, and J. Tani, “Incremental learning of goal- directed actions in a dynamic environment by a robot using active inference,”Entropy, vol. 25, no. 11, p. 1506, 2023

    2023

  31. [39]

    Human–robot kinaesthetic interac- tions based on the free-energy principle,

    H. Sawada, W. Ohata, and J. Tani, “Human–robot kinaesthetic interac- tions based on the free-energy principle,”IEEE Transactions on Systems, Man, and Cybernetics: Systems, 2024

    2024

  32. [40]

    A theory of adaptive pattern classifiers,

    S. Amari, “A theory of adaptive pattern classifiers,”IEEE Transactions on Electronic Computers, no. 3, pp. 299–307, 2006

    2006

  33. [41]

    Adam: A method for stochastic optimization,

    D. P. Kingma and J. Ba, “Adam: A method for stochastic optimization,” arXiv preprint arXiv:1412.6980, 2014

    2014 arXiv

  34. [42]

    Characterizing the sense of agency in human– robot interaction based on the free energy principle,

    W. Ohata and J. Tani, “Characterizing the sense of agency in human– robot interaction based on the free energy principle,”npj Complexity, vol. 2, no. 1, p. 12, 2025

    2025

  35. [43]

    An interpretation of the

    J. Tani, “An interpretation of the ”self” from the dynamical systems per- spective: A constructivist approach,”Journal of Consciousness Studies, vol. 5, no. 5–6, pp. 516–542, 1998

    1998

  36. [44]

    Catastrophic forgetting in connectionist networks: Causes, consequences and solutions,

    R. French, “Catastrophic forgetting in connectionist networks: Causes, consequences and solutions,”Trends in Cognitive Sciences, vol. 3, no. 4, pp. 128–135, 1999

    1999

  37. [45]

    Continual learning with deep generative replay,

    H. Shin, J. K. Lee, J. Kim, and J. Kim, “Continual learning with deep generative replay,”Advances in neural information processing systems, vol. 30, 2017

    2017

  38. [46]

    Catastrophic interference in connec- tionist networks: The sequential learning problem,

    M. McCloskey and N. J. Cohen, “Catastrophic interference in connec- tionist networks: The sequential learning problem,” inPsychology of learning and motivation. Elsevier, 1989, vol. 24, pp. 109–165

    1989

  39. [47]

    Catastrophic forgetting in connectionist networks,

    R. M. French, “Catastrophic forgetting in connectionist networks,” Trends in cognitive sciences, vol. 3, no. 4, pp. 128–135, 1999

    1999

  40. [48]

    R. S. Sutton, A. G. Bartoet al.,Reinforcement learning: An introduction. MIT press Cambridge, 1998, vol. 1, no. 1

    1998

  41. [49]

    Off-policy policy gradient with state distribution correction,

    Y . Liu, A. Swaminathan, A. Agarwal, and E. Brunskill, “Off-policy policy gradient with state distribution correction,”arXiv preprint arXiv:1904.08473, 2019

    1904 arXiv

  42. [50]

    A reduction of imitation learning and structured prediction to no-regret online learning,

    S. Ross, G. Gordon, and D. Bagnell, “A reduction of imitation learning and structured prediction to no-regret online learning,” inProceedings of the fourteenth international conference on artificial intelligence and statistics. JMLR Workshop and Conference Proceedings, 2011, p...

    2011

  43. [51]

    A sensorimotor account of vision and visual consciousness,

    J. K. O’regan and A. No ¨e, “A sensorimotor account of vision and visual consciousness,”Behavioral and brain sciences, vol. 24, no. 5, pp. 939– 973, 2001

    2001

  44. [52]

    Socializing sensorimotor contingencies,

    A. L ¨ubbert, F. G ¨oschl, H. Krause, T. R. Schneider, A. Maye, and A. K. Engel, “Socializing sensorimotor contingencies,”Frontiers in Human Neuroscience, vol. 15, p. 624610, 2021

    2021

  45. [53]

    Sen- sorimotor contingencies as a key drive of development: from babies to robots,

    L. Jacquey, G. Baldassarre, V . G. Santucci, and J. K. O’regan, “Sen- sorimotor contingencies as a key drive of development: from babies to robots,”Frontiers in neurorobotics, vol. 13, p. 98, 2019

    2019

  46. [54]

    On the dynamics of small continuous-time recurrent neural networks,

    R. D. Beer, “On the dynamics of small continuous-time recurrent neural networks,”Adaptive Behavior, vol. 3, no. 4, pp. 469–509, 1995

    1995

  47. [55]

    Auto-encoding variational bayes,

    D. P. Kingma and M. Welling, “Auto-encoding variational bayes,”arXiv preprint arXiv:1312.6114, 2013. 16 SUPPLEMENTARYMATERIAL A. Generation of AI-Tutor Training Trajectories The AI-tutor training trajectories were generated compu- tationally rather than recorded from human dem...

    2013 arXiv

Pith tools

Reviewed June 26, 2026 · model on record in the stance chip above.