Pith. sign in

REVIEW 4 major objections 5 minor 31 references

Decentralized pivoting-cube ensembles can self-reconfigure to target shapes using only locally exchanged information, with near-optimal move counts.

Reviewed by Pith at T0; open to challenge. T0 means a machine referee read the full paper against a public rubric. the ladder, T0–T4 →

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

Decentralized reinforcement learning with local receptive fields can reconfigure pivoting cube ensembles into target shapes, achieving near-optimal move counts with multiple local message-passing rounds.

T0 review reviewed 2026-08-05 challenge →

load-bearing objection A solid empirical study with a real negative result, but the train/test legality mismatch and a loose centralized baseline comparison keep the locality claim from being fully established. the 4 major comments →

arxiv 2509.03140 v1 pith:B6IAANE5 submitted 2025-09-03 cs.NE cs.AIcs.RO

Decentralised self-organisation of pivoting cube ensembles using geometric deep learning

Nadezhda Dobreva , Emmanuel Blazquez , Jai Grover , Dario Izzo , Yuzhen Qin , Dominik Dold This is my paper
classification cs.NE cs.AIcs.RO
keywords modular roboticsself-reconfigurationpivoting cubesdecentralized controlreinforcement learninggeometric deep learninglocal information exchangeconvolutional policy networks
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.

The reading

This paper asks whether a swarm of identical pivoting cubes can reorganize itself into a target shape when no cube sees more than its local neighborhood. It answers yes: a single neural-network policy shared by all cubes, trained by reinforcement learning, succeeds even when each cube can only exchange information with immediate neighbours. Repeated local exchanges effectively accumulate global knowledge, bringing move counts close to optimal—on average about 1.0 to 2.2 pivots per cube depending on the target shape. If this result holds, self-reconfiguring modular hardware does not need a central controller or global communication, which matters for building large structures from many identical units.

Core claim

The central claim is that decentralized control with purely local information is sufficient for reconfiguring homogeneous pivoting cube ensembles in two dimensions. Each cube runs the same convolutional policy, and information is aggregated only within a kernel-defined local radius; nevertheless, trained policies reach the target shape with near-100% success even at radius 1. The paper's quantitative anchor is the best model, which uses two 5×5 mirror-rotation-invariant convolutional layers and needs on average 1.89 moves per cube for the line target, 2.17 for the table, 1.43 for the chair, and 1.01 for the sun-shield—compared with the O(8n) to O(n^2) moves of a centralized line-morphing alg

What carries the argument

The load-bearing mechanism is a fully convolutional policy network shared across all cubes: the binary occupancy image of the ensemble is passed through k×k convolutions, with activations masked to zero in empty cells so information travels only through cubes. Kernel size and number of layers set the local radius—a 3×3 kernel means nearest-neighbour exchange, while stacking layers or using a 5×5 kernel widens the effective reception. A mirror-rotation-invariant variant (MR-CNN) is built from rotation-invariant kernels obtained by projecting weight matrices onto the Reynolds operator, plus paired channels that swap under mirroring, so the policy respects the grid's symmetries by construction.

Load-bearing premise

The load-bearing premise is that the local connectivity check used at test time preserves global connectivity exactly as the full breadth-first check does, so policies trained under the global check still behave as reported under the local check.

What would settle it

Run a trained best model on the same 500 random initial states but evaluate legality with the full breadth-first connectivity search used during training, and compare success rates and move counts to the local-check evaluation; any large drop would show the reported near-100% success depends on the local-check approximation rather than the trained policy.

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

If this is right

  • Decentralized cube ensembles can reconfigure to target shapes with only nearest-neighbour information, removing the need for any cube to see the whole structure.
  • Stacking local information exchanges (more convolution layers) is an effective way to approach global performance while keeping physical interactions local.
  • Move counts near 1–2 pivots per cube beat the best known centralized line-morphing algorithm on the line shape for n=9, suggesting decentralized policies can be competitive in efficiency.
  • Because identical policies can be swapped, an ensemble can morph through different target shapes (e.g., table to chair to line) simply by loading different trained weights.
  • The modest gain from symmetry-aware architectures implies that standard local convolutional networks are a strong baseline for this task, guiding future controllers toward simpler networks.

Where Pith is reading between the lines

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

  • The paper's test-time legality check uses a local connectivity search instead of the full breadth-first search used in training; whether this approximation ever admits moves that break connectivity on configurations reachable by the policy remains a testable question.
  • The one remaining global operation is sampling which cube moves at each timestep; replacing this with asynchronous local voting among nearby cubes would be a natural next step toward a fully asynchronous controller.
  • The authors report that scaling to larger ensembles is an open issue because the search space grows exponentially and curriculum learning, though promising, was unstable; a concrete extension would test whether more local message-passing rounds or improved curricula overcome this.
  • Since the convolutional policy is equivalent to graph message passing, the same training recipe should transfer to 3D cube ensembles or sliding-cube robots by redefining the action space.
Share X Bluesky LinkedIn Reddit HN

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

4 major / 5 minor

Summary. The paper proposes a decentralized control scheme for two-dimensional pivoting cube modular robots, in which every cube is governed by the same neural-network policy trained with masked PPO. The state is encoded as binary and index images, and the policy is a convolutional or mirror-rotation-invariant convolutional network whose receptive field is varied to control locality. Experiments cover four target shapes, several kernel sizes and layer counts, and include training from random initial states, perturbation tests, and target-to-target morphing. The central claim is that even nearest-neighbour information exchange suffices for reconfiguration, while increasing the receptive field improves speed, and that geometric deep learning provides only minor benefits. The paper includes a credible negative result for the geometric inductive biases and a comparison with a centralized pivoting-cube planner for the line target.

Significance. If the central claim holds, the paper is a useful step toward scalable decentralized self-reconfiguration for modular robots, with potential relevance to space systems. The experimental design is thorough in several respects: five training seeds, 500 evaluation states per configuration, perturbation experiments, and multiple target shapes. The negative result that rotation/mirror-invariant convolutions give little benefit is an honest and useful finding. However, the train/test legality mismatch documented in §2.1 is a load-bearing gap: the policies are trained under a global connectivity check and evaluated under a stricter local check, so the reported success rates and move counts may not reflect the competence of the trained policy in the evaluation MDP. In addition, the 'near-optimal' claim in the abstract is not supported by a lower bound, and aggregate results are reported without error bars or significance tests. These issues prevent full confidence in the quantitative conclusions, although the qualitative feasibility result is plausibly sound.

major comments (4)
  1. [§2.1 (last paragraph)] The compromise between full BFS during training and a local connectivity search at test time changes the legal action set between the MDP used for optimization and the MDP used for evaluation. The local check is stricter: a move that preserves global connectivity can fail the local check when the paths connecting the moving cube's neighbours exceed the local radius. Policies trained with action masking on the global-check MDP are therefore evaluated on a different action set, and the reported success rates and move counts (Figs. 4–5) may be artefacts of this mismatch. The paper should either evaluate the same policies under the full BFS check, train policies under the local check from the start, or otherwise explicitly quantify the discrepancy. This is directly load-bearing for the claim that local information exchange is sufficient.
  2. [Abstract/§3 (line comparison)] The abstract states that 'near-optimal reconfiguration' is achieved, but no optimality baseline or lower bound is provided. The comparison in §3 to the centralized algorithm of [15] compares a measured value 1.89n for n=9 against an asymptotic worst-case upper bound O(8n)–O(n^2). This does not establish closeness to the optimal number of moves from random initial states. The term 'near-optimal' should be removed or justified by a proper lower bound or an optimal planner for these small instances.
  3. [§3, Figs. 4–5] All quantitative claims are based on five seeds and aggregate curves with no error bars, confidence intervals, or per-seed breakdowns. Given the paper's own statement about the high variance of RL training, the comparisons among architectures (e.g., 'the best performing model', 'clearly outperforming') are not statistically supported. Please report variance or per-seed values and perform a simple significance test for the main comparisons. In addition, the perturbation-experiment filtering is inconsistent: the text says 'all trials with success rate < 1% are filtered out', while the Fig. 5 caption says 'we only show results for networks that achieved a success rate above 99%'. This filtering can bias the reported move counts and must be clarified.
  4. [§3 and Appendix] The distribution of the '500 random initial states' used for evaluation is not specified: are they sampled uniformly over all connected configurations of N cubes, over configurations reachable within some number of moves from the target, or from the same distribution used during training? Success rates and move counts are highly sensitive to this choice. The paper should define the initial-state sampling procedure explicitly, for both training and evaluation, so that results are reproducible and interpretable.
minor comments (5)
  1. [§2.2.2 / Discussion] The action-sampling step is global and synchronized; this is disclosed, but the abstract and title say 'decentralised' without qualification. The final sentence of the abstract uses 'mostly local control', which is more accurate. Consider making this caveat prominent in the introduction or conclusion.
  2. [§3] The phrase 'moves per cube' is not defined. Is it total moves divided by N, or the average number of times each cube pivots? Please define it explicitly before the first use.
  3. [Appendix] The perturbation experiments are listed as m = {1, 3, 5, 10} in the appendix but the main text says m = {1, 3, 10}. Align the notation. Also, the appendix says '14 repeats' while the main text describes five seeds; clarify the relationship.
  4. [Equation (13)] The reward depends on overlap with the target after applying all 8 dihedral transformations. This means the target shape is defined up to rotation and reflection. This is fine, but it should be stated explicitly in the main text when interpreting 'success', not only in the phase-correlation paragraph.
  5. [§3] In the comparison with [15], the measured constant 1.89n is for n=9 and the asymptotic expression O(8n)–O(n^2) has unknown constants. The word 'clearly outperforming' is too strong without running the centralized algorithm on the same initial states. Soften the wording or add such a comparison.

Circularity Check

0 steps flagged

No circularity: the central results are empirical RL outcomes, not derived quantities; the only caveat is a train/test legality-check mismatch that is a validity concern, not a circular reduction.

full rationale

The paper's central claim—that locally informed policies can reconfigure pivoting cube ensembles—is established by reinforcement learning training and evaluation, not by construction. The reward is a shaped overlap reward with constants α0 = α1 = 0.7 and γ0 = γ1 = 1.2, explicitly described as 'obtained from a parameter sweep'; these are training inputs, not fitted parameters renamed as predictions. The rotation-invariant architecture is adopted from external group-equivariant CNN theory (Cohen & Welling [22]) and the Reynolds operator [23], which are independent mathematical tools. The only self-citations (Izzo & Pettazzi [13,14]) appear in the introduction as examples of satellite-swarm applications and are not load-bearing. The comparison to [15] is an external centralized algorithm, and the reported 1.89n moves per cube is a measured empirical quantity on the authors' own environment, not an equality forced by the reward or architecture. The notable caveat is the Section 2.1 train/test legality-check mismatch: 'we use the full breadth first search during training, while only using a local search – with the same total receptive field size as the local controllers – during test time.' This is a potential validity threat (the test MDP differs from the training MDP), but it is not circularity: the success rates and move counts are empirical outcomes obtained under the stated test protocol, and they are not implied by the training objective or architecture by definition. No circular step is present, so the circularity score is 0.

Axiom & Free-Parameter Ledger

4 free parameters · 4 axioms · 0 invented entities

The central claims do not introduce new physical entities. They rest on simulator fidelity, a graph-theoretic locality claim, standard RL machinery, and hand-tuned reward constants. The reward constants and episode budgets are the main fitted quantities.

free parameters (4)
  • Reward constants alpha_0, alpha_1 = 0.7
    Reward scaling coefficients in Eq. 13, obtained from a parameter sweep (Section 2.4).
  • Reward exponents gamma_0, gamma_1 = 1.2
    Reward shape exponents in Eq. 13, obtained from the same parameter sweep.
  • Episode step budgets per target = 300 (line), 800 (table/chair), 2000 (sun-shield)
    Maximum steps per episode chosen per target shape; directly bounds whether a configuration can be reached and thus affects measured success rates.
  • Layer width configurations = e.g., [1, 8192, 32] for 1-layer; [1, 64, 512, 32] for 2-layer
    Hand-chosen architecture capacities (Appendix); not swept, affecting expressiveness and performance.
axioms (4)
  • domain assumption The pivoting cube dynamics of the SpOC challenge simulator faithfully represent physical pivoting cube robots.
    The legality checks (collision and connectivity) are the only physics in the environment (Section 2.1); no hardware validation is provided.
  • domain assumption Local connectivity of the ensemble after removing a cube implies global connectivity, as long as the ensemble started globally connected.
    Invoked to justify the local connectivity check at test time (Section 2.1).
  • standard math PPO with action masking converges to a good policy for this MDP within the allotted total steps.
    Standard RL assumption; no convergence guarantee specific to this environment is given (Section 2.4).
  • standard math The Reynolds operator and group-equivariant convolution construction is applied correctly to yield rotation-invariant and mirror-alternating kernels.
    The architecture construction (Section 2.3) relies on standard representation theory; no formal verification is provided.

reviewed 2026-08-05 · how reviews work

0 comments
Cite this review

Pith. "Pith review of Decentralised self-organisation of pivoting cube ensembles using geometric deep learning." pith.science (2026). https://pith.science/paper/B6IAANE5

@misc{pith2026250903140,
  author       = {Pith},
  title        = {Pith review of: Decentralised self-organisation of pivoting cube ensembles using geometric deep learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/B6IAANE5}},
  note         = {Machine review of arXiv:2509.03140}
}
Share X Bluesky LinkedIn Reddit HN
read the original abstract

We present a decentralized model for autonomous reconfiguration of homogeneous pivoting cube modular robots in two dimensions. Each cube in the ensemble is controlled by a neural network that only gains information from other cubes in its local neighborhood, trained using reinforcement learning. Furthermore, using geometric deep learning, we include the grid symmetries of the cube ensemble in the neural network architecture. We find that even the most localized versions succeed in reconfiguring to the target shape, although reconfiguration happens faster the more information about the whole ensemble is available to individual cubes. Near-optimal reconfiguration is achieved with only nearest neighbor interactions by using multiple information passing between cubes, allowing them to accumulate more global information about the ensemble. Compared to standard neural network architectures, using geometric deep learning approaches provided only minor benefits. Overall, we successfully demonstrate mostly local control of a modular self-assembling system, which is transferable to other space-relevant systems with different action spaces, such as sliding cube modular robots and CubeSat swarms.

Figures

Figures reproduced from arXiv: 2509.03140 by Dario Izzo, Dominik Dold, Emmanuel Blazquez, Jai Grover, Nadezhda Dobreva, Yuzhen Qin.

Figure 1
Figure 1. Figure 1: Pivoting cube ensembles enable reconfigurable structures. (A) A cube ensemble in two dimensions reconfigures through pivots between two shapes: a table and a chair. (B) Although the dynamics are more complex for microsatellites such as CubeSats, the reconfiguration problem is structurally similar: all cubes are interchangeable, communication between cubes is locally constrained, and they have to reorganize… view at source ↗
Figure 2
Figure 2. Figure 2: Allowed moves and representation of two-dimensional cube ensembles. (A) Only pivots that do not physically collide with other cubes are allowed. (B) Moreover, the ensemble has to stay connected during reconfiguration. (C) In the cube environment, the ensemble is represented as a list of coordinates. From this representation, we extract two image representations: a binary image just encoding the shape of th… view at source ↗
Figure 3
Figure 3. Figure 3: Network architecture. (A) Schematic overview of the model. (B) The two-dimensional grid defining the state space of the cube ensemble has two symmetries. If looking at a single cube (red), rotating the ensemble should not change its probabilities for clockwise (cw) and counter-clockwise (ccw) pivoting. Similarly, when mirroring the cube ensemble, these probabilities should simply swap. (C) A 3×3 and 5×5 co… view at source ↗
Figure 4
Figure 4. Figure 4: Performance for different target shapes and network architectures. We show the success rate for ensembles reconfiguring into different target shapes from any random initial shape, both using ordinary CNNs and MR-CNNs. The local radius indicates from how far away cubes accumulate information to choose their action. For instance, a one layer network with kernel 3 × 3 has a local radius of 1, two layers with … view at source ↗
Figure 5
Figure 5. Figure 5: Average number of moves required for the networks in [PITH_FULL_IMAGE:figures/full_fig_p008_5.png] view at source ↗
Figure 6
Figure 6. Figure 6: Reconfiguring between different target shapes. Here, we initialize the network in the table shape and let it reconfigure to a chair. Start End [PITH_FULL_IMAGE:figures/full_fig_p009_6.png] view at source ↗
Figure 7
Figure 7. Figure 7: Reconfiguring between different target shapes. Here, we initialize the network in the chair shape and let it reconfigure to a line. Lastly, since the models learned to reconfigure into their target shapes from any initial state, this means we can morph the ensemble between target shapes by swapping out the network parameters. This is illustrated in [PITH_FULL_IMAGE:figures/full_fig_p009_7.png] view at source ↗

discussion (0)

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

Reference graph

Works this paper leans on

31 extracted references · 28 canonical work pages · 2 internal anchors

  1. [1]

    Self assembly in space using behaviour based intelligent components

    M. Ayre, D. Izzo, and L. Pettazzi. “Self assembly in space using behaviour based intelligent components”. In: TAROS, Towards Autonomous Robotic Systems(2005)

    work page 2005

  2. [2]

    Using CubeSat/micro-satellite technology to demonstrate the Autonomous Assembly of a Reconfigurable Space Telescope (AAReST)

    C. Underwood et al. “Using CubeSat/micro-satellite technology to demonstrate the Autonomous Assembly of a Reconfigurable Space Telescope (AAReST)”. In: Acta Astronautica 114 (2015), pp. 112–122

    work page 2015

  3. [3]

    Architecture for in-space robotic assembly of a modular space telescope

    N. Lee et al. “Architecture for in-space robotic assembly of a modular space telescope”. In: Journal of Astro- nomical Telescopes, Instruments, and Systems 2.4 (2016), pp. 041207–041207. 10

    work page 2016

  4. [4]

    TESSERAE: Self-assembling shell structures for space exploration

    A. Ekblaw and J. Paradiso. “TESSERAE: Self-assembling shell structures for space exploration”. In: Proceed- ings of IASS Annual Symposia. V ol. 2018. 1. International Association for Shell and Spatial Structures (IASS). 2018, pp. 1–8

    work page 2018

  5. [5]

    ElectroV oxel: Electromagnetically actuated pivoting for scalable modular self-reconfigurable robots

    M. Nisser et al. “ElectroV oxel: Electromagnetically actuated pivoting for scalable modular self-reconfigurable robots”. In: 2022 International Conference on Robotics and Automation (ICRA). IEEE. 2022, pp. 4254–4260

    work page 2022

  6. [6]

    Toward the autonomous assembly of large telescopes using CubeSat rendezvous and docking

    C. Pirat et al. “Toward the autonomous assembly of large telescopes using CubeSat rendezvous and docking”. In: Journal of Spacecraft and Rockets 59.2 (2022), pp. 375–388

    work page 2022

  7. [7]

    The science performance of JWST as characterized in commissioning

    J. Rigby et al. “The science performance of JWST as characterized in commissioning”. In: Publications of the Astronomical Society of the Pacific135.1046 (2023), p. 048001

    work page 2023

  8. [8]

    Meshing complex macro-scale objects into self-assembling bricks

    A. Hacohen et al. “Meshing complex macro-scale objects into self-assembling bricks”. In: Scientific reports 5.1 (2015), p. 12257

    work page 2015

  9. [9]

    Recent Developments in Self-Assembling Multi-Robot Systems

    E. Bray and R. Groß. “Recent Developments in Self-Assembling Multi-Robot Systems”. In: Current Robotics Reports 4.4 (2023), pp. 101–116

    work page 2023

  10. [10]

    Oscillators that sync and swarm

    K. P. O’Keeffe, H. Hong, and S. H. Strogatz. “Oscillators that sync and swarm”. In:Nature communications 8.1 (2017), p. 1504

    work page 2017

  11. [11]

    Growing neural cellular automata

    A. Mordvintsev et al. “Growing neural cellular automata”. In: Distill 5.2 (2020), e23

    work page 2020

  12. [12]

    Insect-inspired AI for autonomous robots

    G. C. de Croon et al. “Insect-inspired AI for autonomous robots”. In: Science robotics 7.67 (2022), eabl6334

    work page 2022

  13. [13]

    Equilibrium shaping: distributed motion planning for satellite swarm

    D. Izzo and L. Pettazzi. “Equilibrium shaping: distributed motion planning for satellite swarm”. In: Proc. 8th Intern. Symp. on Artificial Intelligence, Robotics and Automation in space. V ol. 25. 2005

    work page 2005

  14. [14]

    Autonomous and distributed motion planning for satellite swarm

    D. Izzo and L. Pettazzi. “Autonomous and distributed motion planning for satellite swarm”. In: Journal of Guidance, Control, and Dynamics 30.2 (2007), pp. 449–459

    work page 2007

  15. [15]

    Reconfiguration planning for pivoting cube modular robots

    C. Sung et al. “Reconfiguration planning for pivoting cube modular robots”. In: 2015 IEEE international conference on robotics and automation (ICRA). IEEE. 2015, pp. 1933–1940

    work page 2015

  16. [16]

    Autonomous reconfiguration of homogeneous pivoting cube modular satellite by deep reinforce- ment learning

    Q. Song et al. “Autonomous reconfiguration of homogeneous pivoting cube modular satellite by deep reinforce- ment learning”. In: Proceedings of the Institution of Mechanical Engineers, Part I: Journal of Systems and Control Engineering 235.10 (2021), pp. 1777–1786

    work page 2021

  17. [17]

    Reinforcement-learning-based task planning for self-reconfiguration of cellular satellites

    Y . Zhang et al. “Reinforcement-learning-based task planning for self-reconfiguration of cellular satellites”. In: IEEE Aerospace and Electronic Systems Magazine 37.6 (2021), pp. 38–47

    work page 2021

  18. [18]

    Proximal policy optimization algorithms

    J. Schulman et al. “Proximal policy optimization algorithms”. In: arXiv preprint arXiv:1707.06347 (2017)

    Pith/arXiv arXiv 2017

  19. [19]

    The Space Optimization Competition

    M. Bannach et al. “The Space Optimization Competition”. In: Proceedings of the Genetic and Evolutionary Computation Conference Companion. 2024, pp. 21–22

    work page 2024

  20. [20]

    Openai gym

    G. Brockman et al. “Openai gym”. In: arXiv preprint arXiv:1606.01540 (2016)

    Pith/arXiv arXiv 2016

  21. [21]

    Geometric deep learning: going beyond euclidean data

    M. M. Bronstein et al. “Geometric deep learning: going beyond euclidean data”. In: IEEE Signal Processing Magazine 34.4 (2017), pp. 18–42

    work page 2017

  22. [22]

    Group equivariant convolutional networks

    T. Cohen and M. Welling. “Group equivariant convolutional networks”. In:International conference on machine learning. PMLR. 2016, pp. 2990–2999

    work page 2016

  23. [23]

    Mumford, J

    D. Mumford, J. Fogarty, and F. Kirwan. Geometric invariant theory. V ol. 34. Springer Science & Business Media, 1994

    work page 1994

  24. [24]

    Neural Networks for Learning Counterfactual G-Invariances from Single Environments

    S. C. Mouli and B. Ribeiro. “Neural networks for learning counterfactual g-invariances from single environ- ments”. In: arXiv preprint arXiv:2104.10105 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  25. [25]

    Deep q-learning versus proximal policy optimization: Performance comparison in a material sorting task

    R. Kozlica, S. Wegenkittl, and S. Hiränder. “Deep q-learning versus proximal policy optimization: Performance comparison in a material sorting task”. In: 2023 IEEE 32nd International Symposium on Industrial Electronics (ISIE). IEEE. 2023, pp. 1–6

    work page 2023

  26. [26]

    The phase correlation image alignment method

    C. D. Kuglin. “The phase correlation image alignment method”. In: IEEE Int. Conf. on Cybernetics and Society,

    work page

  27. [27]

    Efficient subpixel image registration algorithms

    M. Guizar-Sicairos, S. T. Thurman, and J. R. Fienup. “Efficient subpixel image registration algorithms”. In: Optics letters 33.2 (2008), pp. 156–158

    work page 2008

  28. [28]

    Deep reinforcement learning that matters

    P. Henderson et al. “Deep reinforcement learning that matters”. In: Proceedings of the AAAI conference on artificial intelligence. V ol. 32. 1. 2018. 11

    work page 2018

  29. [29]

    A survey on curriculum learning

    X. Wang, Y . Chen, and W. Zhu. “A survey on curriculum learning”. In:IEEE transactions on pattern analysis and machine intelligence 44.9 (2021), pp. 4555–4576

    work page 2021

  30. [30]

    Graph attention networks

    P. Veli ˇckovi´c et al. “Graph attention networks”. In: arXiv preprint arXiv:1710.10903 (2017)

    Pith/arXiv arXiv 2017

  31. [31]

    Reconfiguration Algorithms for Cubic Modular Robots with Realistic Movement Constraints

    N. S. R. Team et al. “Reconfiguration Algorithms for Cubic Modular Robots with Realistic Movement Con- straints”. In: arXiv preprint arXiv:2405.15724 (2024). Appendix This appendix provides further information on our experiment setup. All experiments are conducted 5 times, with different random seeds: 12345, 32823, 57923, 70852, 97245. The trained network...

    work page internal anchor Pith review Pith/arXiv arXiv 2024

This paper was first reviewed by deepseek-v4-flash on August 5, 2026.