Pith. sign in

REVIEW 3 major objections 5 minor 60 references

Local-Canonicalization Equivariant Graph Neural Networks for Sample-Efficient and Generalizable Swarm Robot Control

T0 review · 3 major / 5 minor · reviewed 2026-08-15 · deepseek-v4-flash

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

Pith's one-line read Canonicalizing each agent's observation before graph encoding yields an E(2)-equivariant policy that transfers across team sizes.

desk verdict A useful, well-ablationed MARL architecture whose competitive-results claims hang on an unresolved global-vs-local center-of-mass ambiguity under partial observability. read the letter →

arxiv 2509.14431 v2 pith:EOBLTWW5 submitted 2025-09-17 cs.RO

Keqin Wang , Tao Zhong , David Chang , Christine Allen-Blanchette This is my paper · ORCID
classification cs.RO
keywords multi-agentreinforcementlearningequivariantpolicycanonicalizationgraphneuralnetworksswarmcontrolzero-shotgeneralizationE(2)equivariancerole-awareencoding
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 introduces LEGO, a policy architecture for multi-agent reinforcement learning, and claims that combining per-agent canonicalization with role-aware graph encoding makes learned swarm-control policies equivariant to the Euclidean symmetries of the plane and permutation-symmetric within roles. The practical claim is that with these symmetries built in, policies train much more sample-efficiently and, unlike standard MLP policies, keep working when the team grows or shrinks, when the scene is rotated or shifted, and when an agent fails. The paper supports this with MAPPO-instantiated experiments in cooperative coverage and competitive pursuit-evasion with occlusion, plus quadrotor trials.

What carries the argument

The load-bearing object is the agent-centric canonical frame $R_i$ built from the agent's velocity $v_i$ and the team center of mass $c = \frac{1}{N}\sum_j p_j$, which maps the global state $X$ into an E(2)-invariant local observation $C_i(X) = \rho_{SE(2)}(g_i^{-1})X$. This frame does two jobs: it removes global rotation and translation as nuisance variation before learning, and, because the predicted local action is multiplied back by $R_i$, it guarantees the end-to-end policy is E(2)-equivariant. The second component is the role-aware Graphormer, a transformer-style attention encoder applied separately to dense subgraphs of each role and then pooled per role, which supplies permutation equivariance within roles and a fixed-size representation across team sizes.

What would settle it

Train the same LEGO-MAPPO architecture on the occluded Tag task twice, once with the global center of mass available for canonicalization and once with the frame built only from visible agents; if the second variant's reward or the E(2)-invariance of its actions degrades substantially, the claim of decentralized E(2)-equivariant policies under partial observability is not supported. A second check is to rotate and translate the initial configuration by a known E(2) transformation while preserving relative geometry, then verify that the policy's rendered global actions rotate and translate identically at every timestep.

Watch

Extended reading notes

Core claim

The central claim is that LEGO produces an E(2)-equivariant policy by construction: each agent first rewrites the global scene in its own local frame, whose x-axis is the agent's velocity and whose y-axis is oriented toward the team's center of mass; a role-aware graph transformer then encodes the canonicalized scene; the actor outputs a local action; and rotating that local action back by the same frame yields a global action that transforms correctly under rotation and translation of the whole scene. Because the encoder's output is pooled per role rather than per agent, the representation's dimension depends on the number of roles, not the number of agents, so the same policy can act for swarms of different sizes. The paper's experiments claim this yields faster convergence and higher rewards than MLP-based, graph-only, and canonicalization-only baselines, zero-shot transfer to 2 through 6 agents, out-of-distribution robustness, and continued operation after one pursuer breaks down in real-world trials.

Load-bearing premise

The canonical y-axis is defined from the global center of mass of all agents, and the claimed invariance and decentralization of the policy hold only if each agent can compute that center of mass from its own observation; under occlusion the paper does not state how.

Editorial extensions

If this is right

  • The same policy trained on one team size can be deployed, without fine-tuning, on smaller and larger swarms; the paper reports this for a 4-agent policy evaluated on 2, 3, 5, and 6 agents.
  • Euclidean equivariance removes the need to learn rotations and translations from data; the reported training curves converge faster and reach higher rewards than MLP, canonicalization-only, and GNN-only baselines.
  • Role-wise pooling keeps the representation size fixed as the team grows, so the architecture can be used as a drop-in actor for MAPPO and is claimed to integrate with standard MARL algorithms.
  • Because the policy tolerates agent removal, including one pursuer disabled mid-episode in real-world trials, it offers graceful degradation for swarm missions.
  • Curriculum initialization from a smaller team gives additional gains on larger teams, with the paper reporting best results when a 4-agent policy warm-starts training on 6, 7, and 8 agents.

Reading between the lines

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

  • A testable extension the paper does not report: building the canonical frame from only the visible subset of agents in the occluded task would determine whether the policy remains decentralized and E(2)-invariant when the global center of mass is unavailable.
  • Because each role is encoded in a separate subgraph and pooled independently, the architecture could plausibly admit an entirely new role at deployment by adding one more role-stream; the paper does not demonstrate this.
  • The curriculum result suggests that the warm-start benefit comes partly from reusing the canonicalization geometry itself, so the gain may grow with team size; this is an inference, not a claim the paper makes.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper introduces LEGO, a modular policy architecture for multi-agent reinforcement learning that combines agent-centric canonicalization (to achieve E(2) equivariance), role-based Graphormer encoders (for permutation equivariance and size generalization), and MAPPO as the training algorithm. The authors claim improved sample efficiency and performance over MLP-based, graph-only, and canonicalization-only baselines on the cooperative MPE Spread task and the competitive Tag-occlusion task, zero-shot generalization to unseen team sizes, robustness to spatial distribution shifts, and a real-world Crazyflie demonstration after an agent failure. The equivariance property is argued from the structure of the canonicalization-decanonicalization pipeline.

Significance. If the claims hold, LEGO offers a practical and modular way to inject geometric and permutation symmetries into standard MARL algorithms, which is valuable for swarm control. Strengths of the paper include a clean architectural decomposition, a public code repository, comparison against several natural baselines, and real-hardware validation with graceful degradation after agent failure. However, the central formal claim of E(2)-equivariant policies is not fully established for the partially observable Tag-occlusion task because the canonicalization relies on the global center of mass, which may not be available under occlusion. The absence of a true equivariant MARL baseline also tempers the comparative claims.

major comments (3)
  1. [IV-B, Eq. (2); V-B] In Eq. (2), the canonical y-axis is defined using the global center of mass c = (1/N) Σ_j p_j over all N agents. In the Tag-occlusion task (Section V-B), each agent observes only teammates and opponents that are not occluded, so this c cannot be computed from O_i. The paper does not state whether c is formed from the full global state X or from the visible subset. If the full state is used at execution, the actor is not decentralized under CTDE and the partial-observability benchmark is not clean; if only visible agents are used, Eq. (2) is not implemented as written and the claimed E(2)-invariance must be re-derived for a visibility-dependent CoM, which is not guaranteed to hold. This ambiguity is load-bearing for the competitive Tag-occlusion results and for the paper's 'local canonicalization' contribution.
  2. [IV-B/E and III] The formal equivariance statement in Section IV-B, C_i(ρ_E(2)(R,t) X) = C_i(X), is stated for the global state X, but the deployed policy acts on observations O_i of a partially observable Markov game (Section III). In Tag-occlusion, the observation function is defined through occlusion by obstacles, and the paper does not specify a group action on O_i that commutes with the model. Without such a specification, the 'by construction' E(2)-equivariance of the policy does not directly extend to the actual observation-to-action mapping. Please state the assumptions under which policy equivariance holds, or restrict the claim to fully observed settings.
  3. [V-A and Abstract] The abstract claims that LEGO outperforms 'equivariant baselines,' but the experimental comparison in Section V-A includes MAPPO, MAPPO-local, and MAPPO-GNN; no true equivariant MARL method from the related-work list (e.g., [39] or [42]) is evaluated. The MAPPO-local ablation is useful, but it does not substantiate superiority over existing equivariant architectures. Please add a stronger equivariant baseline or soften the wording.
minor comments (5)
  1. [IV-D] The role-wise pooling is described as pooling over all nodes in each subgraph, which would yield the same pooled vector for every agent of the same role; please clarify whether the agent's own node is treated separately or whether agent identity is otherwise preserved, since this affects how same-role agents differentiate.
  2. [Fig. 2 caption and IV-B] The caption says the 'global observation X' is canonicalized, while the text defines the canonicalization on X as the global state; please align the notation between global state and agent observation to avoid confusion about what is available at execution.
  3. [Introduction and Abstract] The introduction and abstract use E(n) while Section IV-B defines E(2); please use the dimension consistently.
  4. [Table I] The dagger symbol is used for MAPPO, MAPPO-GNN, and LEGO-MAPPO at 4 agents to indicate models specially trained for that configuration; please check whether the markers apply to MAPPO-GNN as intended, given the text states only LEGO is trained at 4 agents.
  5. [V-D] The real-world experiments use the global state X to track positions; please clarify whether the deployed actors receive local observations or full state, to reconcile with the partial-observability claims made elsewhere.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: LEGO's equivariance is a structural construction, and all performance claims are benchmarked against external baselines without fitted parameters or self-citation chains.

full rationale

The paper's central derivation is the E(2)-equivariance of the policy: Eq. 3 canonicalizes the global state X into the agent-centric frame C_i(X), and Eq. 6 rotates the local action back with R_i. This is an algebraic identity (C_i(rho X) = C_i(X), and the global action transforms as a_i' = R_i' a_loc = g R_i a_loc), not a fitted or predicted quantity. No parameter is fit to the evaluation targets and then renamed as a prediction: the sample-efficiency curves, zero-shot scalability results, and out-of-distribution results are direct benchmark comparisons against MAPPO, MAPPO-local, and MAPPO-GNN. The self-citations are background references for GNN and equivariance techniques and are not load-bearing; no uniqueness theorem or ansatz is imported from them. The only noteworthy tension is that Eq. 2's center of mass uses all N agents while the Tag-occlusion task is partially observable, but this is a correctness/decentralization concern rather than circularity: it does not reduce any claim to its own inputs. Under the stated rules, the honest finding is no significant circularity.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central equivariance construction introduces no fitted physical constants; the symmetry handling is parameter-free. The main unstated assumptions concern observability (global CoM under occlusion), the well-definedness of the canonical frame at degenerate configurations, and the match between E(2) equivariance and the actual MDP symmetries. No invented physical entities are introduced.

free parameters (1)
  • Curriculum pretraining budget = 1e6 environment steps on the 4-agent Spread task
    Section V-C uses a hand-chosen warm-start budget of 1e6 steps followed by 4e6 target-training steps. This is an experimental hyperparameter, not a fitted physical constant, but it is part of the curriculum learning result.
assumptions (5)
  • domain assumption Global center of mass c = (1/N) sum_j p_j is available to every agent at execution time.
    Section IV-B Eq. 2 defines c over all N agents to build the canonical frame, but Section V-A states Tag-occlusion agents only observe non-occluded peers. The paper does not explain how CoM is computed under occlusion.
  • domain assumption The canonical frame is well-defined and continuous except at the stated v_i = 0 degenerate point.
    Eq. 1 handles v_i = 0, but the sign term in Eq. 2 is undefined when x_i^T J d_i = 0, i.e., when the agent, its velocity, and the team center are collinear. No treatment of this discontinuity is given, which can break smoothness assumptions used by PPO.
  • domain assumption The MDP itself is E(2)-symmetric, so an equivariant policy class contains an optimal policy.
    Section IV-E assumes E(2)-equivariance is beneficial, but map boundaries and out-of-bounds penalties in Tag-occlusion break exact translation symmetry if boundary information is not part of the observation. The paper does not test or discuss this mismatch.
  • domain assumption Role-wise pooling over Graphormer encodings preserves enough information for optimal action selection.
    Section IV-D collapses each role subgraph into a fixed vector. This enables zero-shot scaling but removes per-agent identity and can lose relational details such as which specific evader is closest unless preserved by the pooling statistics.
  • standard math Graph attention with shared weights and permutation-invariant pooling is permutation equivariant.
    Standard property of attention and pooling, used in Section IV-D to claim intra-role permutation equivariance.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Local-Canonicalization Equivariant Graph Neural Networks for Sample-Efficient and Generalizable Swarm Robot Control." pith.science (2026). https://pith.science/paper/EOBLTWW5

@misc{pith2026250914431,
  author       = {Pith},
  title        = {Pith review of: Local-Canonicalization Equivariant Graph Neural Networks for Sample-Efficient and Generalizable Swarm Robot Control},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/EOBLTWW5}},
  note         = {Machine review of arXiv:2509.14431}
}
read the original abstract

Multi-agent reinforcement learning (MARL) policies for swarm control often learn inefficiently and generalize poorly across coordinate frames, team sizes, and agent roles. We introduce Local-Canonicalization Equivariant Graph Neural Networks (LEGO), a modular policy architecture that combines agent-centric canonicalization with role-aware graph encoding. Canonicalization removes dependence on the global coordinate frame, while transforming predicted local actions back to the world frame produces an E(2)-equivariant policy. Role-wise graph encoders provide intra-role permutation equivariance and fixed-dimensional representations for variable-size teams. LEGO can be paired with standard MARL algorithms; we instantiate it with MAPPO. Across cooperative MPE Spread and competitive Tag-occlusion benchmarks, LEGO-MAPPO improves sample efficiency and task performance relative to MLP-based, graph-only, canonicalization-only, and equivariant baselines. The learned policies transfer without fine-tuning to unseen team sizes, maintain performance under spatial distribution shifts, and benefit from curriculum initialization for larger teams. In Crazyflie experiments, the policy remains operational after one pursuer is disabled. Code is available at https://github.com/CAB-Lab-Princeton/LEGO-MARL.

Figures

Figures reproduced from arXiv: 2509.14431 by the authors.

Figure 1
Figure 1. An example of equivariance in the MPE Tag environment where pursuers chase evaders while passing though obstacles. (Middle to left) As the agents (circles) are permuted by swapping their indices (s ∈ S(2)), the optimal actions (arrows) are permuted in the same way. (Middle to right) As the agent positions are rotated 90◦ (g ∈ SO(2)), the optimal actions are also rotated. exploit the inherent geometric symmetries pre… view at source ↗
Figure 2
Figure 2. An example of using LEGO in MARL. For each agent i (the upper left red agent here), its global observation X is canonicalized into the local frame Ci(X) = {v ′ i , ρ(g −1 i )X}. Role-based subgraphs (e.g., self, pursuers, evaders, obstacles) are encoded with Graphomer, pooled by role, and concatenated into X˜. The policy and value networks then output the local action and value as a loc i ∼ π(X˜), Vi = f(X˜), with t… view at source ↗
Figure 3
Figure 3. Comparing learning performance on MPE Spread and Tag-occlusion tasks. (A) Average episode rewards in the MPE Spread task with 3 agents and 3 landmarks. (B) Average episode rewards in the MPE Spread task with 6 agents and 6 landmarks. (C) Average episode rewards in the Tag-occlusion task with 2 evaders, 3 pursuers, and 2 obstacles. (D) Average episode rewards in the cross-validation task. centric coordinate frame des… view at source ↗
Figures from the paper (4 more)
Figure 4
Figure 4. Figure 4: Trajectories controlled by policies trained with LEGO￾MAPPO and MAPPO at different training stages. (A) LEGO￾MAPPO after 1×106 interactions. (B) LEGO-MAPPO after 5×106 interactions. (C) MAPPO after 5 × 106 interactions. setting is presented in [PITH_FULL_IMAGE:figures…
Figure 5
Figure 5. Figure 5: Example trajectories in the cross-validation setting. (A) MAPPO-controlled evaders are chased by LEGO-MAPPO￾controlled pursuers. (B) A LEGO-MAPPO-controlled evader is chased by MAPPO-controlled pursuers. the opponent trained with MAPPO, thereby isolating and assessing …
Figure 6
Figure 6. Figure 6: The illustration of the out-of-distribution task, where the agents (blue circle) need to cover the landmarks (black circle) without collision. TABLE III: Out-of-distribution reward result. Method Training initialization Testing initialization left-side right-side unifo…
Figure 8
Figure 8. Figure 8: Trajectories from three real-world trials demonstrating robustness. × indicates the point where one pursuer intentionally breaks down, after which the remaining pursuer successfully adapts its strategy to continue chasing the evader. with heterogeneous role-based graph…

Review history (2 revisions) →

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

60 extracted references · 25 canonical work pages

  1. [39]

    E(3)-equivariant actor-critic methods for cooperative multi-agent reinforcement learning,

    D. Chen and Q. Zhang, “E(3)-equivariant actor-critic methods for cooperative multi-agent reinforcement learning,”arXiv preprint arXiv:2308.11842, 2023

    arXiv 2023

  2. [51]

    Boost- ing sample efficiency and generalization in multi-agent reinforcement learning via equivariance,

    J. McClellan, N. Haghani, J. Winder, F. Huang, and P. Tokekar, “Boost- ing sample efficiency and generalization in multi-agent reinforcement learning via equivariance,”Advances in Neural Information Processing Systems, vol. 37, pp. 41 132–41 156, 2024

    work page 2024

  3. [42]

    Penguin: Partially equivariant graph neural networks for sample efficient marl,

    J. McClellan, G. Brothers, F. Huang, and P. Tokekar, “Penguin: Partially equivariant graph neural networks for sample efficient marl,” arXiv preprint arXiv:2503.15615, 2025

    arXiv 2025

  4. [1]

    A survey on large-population systems and scalable multi- agent reinforcement learning,

    K. Cui, A. Tahir, G. Ekinci, A. Elshamanhory, Y . Eich, M. Li, and H. Koeppl, “A survey on large-population systems and scalable multi- agent reinforcement learning,”arXiv preprint arXiv:2209.03859, 2022

    arXiv 2022

  5. [2]

    A comprehensive survey of multiagent reinforcement learning,

    L. Busoniu, R. Babuska, and B. De Schutter, “A comprehensive survey of multiagent reinforcement learning,”IEEE Transactions on Systems, Man, and Cybernetics, Part C (Applications and Reviews), vol. 38, no. 2, pp. 156–172, 2008

    2008

  6. [3]

    Multi-agent reinforcement learning: A comprehensive survey,

    D. Huh and P. Mohapatra, “Multi-agent reinforcement learning: A comprehensive survey,”arXiv preprint arXiv:2312.10256, 2023

    arXiv 2023

  7. [4]

    Multi-agent reinforcement learning: A review of challenges and applications,

    L. Canese, G. C. Cardarilli, L. Di Nunzio, R. Fazzolari, D. Giardino, M. Re, and S. Spanò, “Multi-agent reinforcement learning: A review of challenges and applications,”Applied Sciences, vol. 11, no. 11, p. 4948, 2021

    work page 2021

  8. [5]

    Coach-player multi-agent reinforcement learning for dynamic team composition,

    B. Liu, Q. Liu, P. Stone, A. Garg, Y . Zhu, and A. Anandkumar, “Coach-player multi-agent reinforcement learning for dynamic team composition,” inInternational Conference on Machine Learning. PMLR, 2021, pp. 6860–6870

    work page 2021

Show all 60 references
  1. [6]

    Learning transferable coopera- tive behavior in multi-agent teams,

    A. Agarwal, S. Kumar, and K. Sycara, “Learning transferable coopera- tive behavior in multi-agent teams,”arXiv preprint arXiv:1906.01202, 2019

    1906 arXiv

  2. [7]

    A multitask- based transfer framework for cooperative multi-agent reinforcement learning,

    C. Hu, C. Wang, W. Luo, C. Yang, L. Xiang, and Z. He, “A multitask- based transfer framework for cooperative multi-agent reinforcement learning,”Applied Sciences, vol. 15, no. 4, p. 2216, 2025

    2025

  3. [8]

    Mdp homomorphic networks: Group symmetries in reinforcement learning,

    E. Van der Pol, D. Worrall, H. van Hoof, F. Oliehoek, and M. Welling, “Mdp homomorphic networks: Group symmetries in reinforcement learning,”Advances in Neural Information Processing Systems, vol. 33, pp. 4199–4210, 2020

    2020

  4. [9]

    so(2)-equivariant reinforcement learning,

    D. Wang, R. Walters, and R. Platt, “so(2)-equivariant reinforcement learning,”arXiv preprint arXiv:2203.04439, 2022

    2022 arXiv

  5. [10]

    A survey of scalable re- inforcement learning,

    G. B. Stone, D. A. Talbert, and W. Eberle, “A survey of scalable re- inforcement learning,”International Journal of Intelligent Computing Research, vol. 13, pp. 1118–1124, 2022

    2022

  6. [11]

    Equivariant reinforcement learning under partial observability,

    H. H. Nguyen, A. Baisero, D. Klee, D. Wang, R. Platt, and C. Amato, “Equivariant reinforcement learning under partial observability,” in Conference on Robot Learning. PMLR, 2023, pp. 3309–3320

    2023

  7. [12]

    Roma: Multi- agent reinforcement learning with emergent roles,

    T. Wang, H. Dong, V . Lesser, and C. Zhang, “Roma: Multi- agent reinforcement learning with emergent roles,”arXiv preprint arXiv:2003.08039, 2020

    2003 arXiv

  8. [13]

    Heterogeneous-agent reinforcement learning,

    Y . Zhong, J. G. Kuba, X. Feng, S. Hu, J. Ji, and Y . Yang, “Heterogeneous-agent reinforcement learning,”Journal of Machine Learning Research, vol. 25, no. 32, pp. 1–67, 2024

    2024

  9. [14]

    The surprising effectiveness of ppo in cooperative multi-agent games,

    C. Yu, A. Velu, E. Vinitsky, J. Gao, Y . Wang, A. Bayen, and Y . Wu, “The surprising effectiveness of ppo in cooperative multi-agent games,”Advances in neural information processing systems, vol. 35, pp. 24 611–24 624, 2022

    2022

  10. [15]

    A comprehensive survey on graph neural networks,

    Z. Wu, S. Pan, F. Chen, G. Long, C. Zhang, and P. S. Yu, “A comprehensive survey on graph neural networks,”IEEE transactions on neural networks and learning systems, vol. 32, no. 1, pp. 4–24, 2020

    2020

  11. [16]

    Attending to graph transformers,

    L. Müller, M. Galkin, C. Morris, and L. Rampášek, “Attending to graph transformers,”arXiv preprint arXiv:2302.04181, 2023

    2023 arXiv

  12. [17]

    Markov games as a framework for multi-agent rein- forcement learning,

    M. L. Littman, “Markov games as a framework for multi-agent rein- forcement learning,” inMachine learning proceedings 1994. Elsevier, 1994, pp. 157–163

    1994

  13. [18]

    Multi-agent reinforcement learning: Independent vs. cooper- ative agents,

    M. Tan, “Multi-agent reinforcement learning: Independent vs. cooper- ative agents,” inProceedings of the tenth international conference on machine learning, 1993, pp. 330–337

    1993

  14. [19]

    Deep decentralized multi-task multi-agent reinforcement learning under par- tial observability,

    S. Omidshafiei, J. Pazis, C. Amato, J. P. How, and J. Vian, “Deep decentralized multi-task multi-agent reinforcement learning under par- tial observability,” inInternational conference on machine learning. PMLR, 2017, pp. 2681–2690

    2017

  15. [20]

    F. A. Oliehoek, C. Amatoet al.,A concise introduction to decentral- ized POMDPs. Springer, 2016, vol. 1

    2016

  16. [21]

    Value-decomposition networks for cooperative multi-agent learning,

    P. Sunehag, G. Lever, A. Gruslys, W. M. Czarnecki, V . Zambaldi, M. Jaderberg, M. Lanctot, N. Sonnerat, J. Z. Leibo, K. Tuylset al., “Value-decomposition networks for cooperative multi-agent learning,” arXiv preprint arXiv:1706.05296, 2017

    2017 arXiv

  17. [22]

    Monotonic value function factorisation for deep multi- agent reinforcement learning,

    T. Rashid, M. Samvelyan, C. S. De Witt, G. Farquhar, J. Foerster, and S. Whiteson, “Monotonic value function factorisation for deep multi- agent reinforcement learning,”Journal of Machine Learning Research, vol. 21, no. 178, pp. 1–51, 2020

    2020

  18. [23]

    Multi-agent actor-critic for mixed cooperative-competitive environments,

    R. Lowe, Y . I. Wu, A. Tamar, J. Harb, O. Pieter Abbeel, and I. Mor- datch, “Multi-agent actor-critic for mixed cooperative-competitive environments,”Advances in neural information processing systems, vol. 30, 2017

    2017

  19. [24]

    Transfer learning in multi-agent reinforcement learning domains,

    G. Boutsioukis, I. Partalas, and I. Vlahavas, “Transfer learning in multi-agent reinforcement learning domains,” inEuropean workshop on reinforcement learning. Springer, 2011, pp. 249–260

    2011

  20. [25]

    Graph convolutional neural networks for web-scale recommender systems,

    R. Ying, R. He, K. Chen, P. Eksombatchai, W. L. Hamilton, and J. Leskovec, “Graph convolutional neural networks for web-scale recommender systems,” inProceedings of the 24th ACM SIGKDD international conference on knowledge discovery & data mining, 2018, pp. 974–983

    2018

  21. [26]

    Molecular contrastive learning of representations via graph neural networks,

    Y . Wang, J. Wang, Z. Cao, and A. Barati Farimani, “Molecular contrastive learning of representations via graph neural networks,” Nature Machine Intelligence, vol. 4, no. 3, pp. 279–287, 2022

    2022

  22. [27]

    Prediction of protein–protein interac- tion using graph neural networks,

    K. Jha, S. Saha, and H. Singh, “Prediction of protein–protein interac- tion using graph neural networks,”Scientific Reports, vol. 12, no. 1, p. 8360, 2022

    2022

  23. [28]

    Deeprank-gnn: a graph neural network framework to learn patterns in protein–protein interfaces,

    M. Réau, N. Renaud, L. C. Xue, and A. M. Bonvin, “Deeprank-gnn: a graph neural network framework to learn patterns in protein–protein interfaces,”Bioinformatics, vol. 39, no. 1, 2023

    2023

  24. [29]

    Behavior-inspired neural networks for relational inference,

    Y . Yang, B. Feng, K. Wang, N. E. Leonard, A. B. Dieng, and C. Allen-Blanchette, “Behavior-inspired neural networks for relational inference,”arXiv preprint arXiv:2406.14746, 2024

    2024 arXiv

  25. [30]

    Resolving over- smoothing with opinion dissensus,

    K. Wang, Y . Yang, I. Saha, and C. Allen-Blanchette, “Resolving over- smoothing with opinion dissensus,”arXiv preprint arXiv:2501.19089, 2025

    2025 arXiv

  26. [31]

    Qmix-gnn: A graph neural network- based heterogeneous multi-agent reinforcement learning model for im- proved collaboration and decision-making,

    T. Zhao, T. Chen, and B. Zhang, “Qmix-gnn: A graph neural network- based heterogeneous multi-agent reinforcement learning model for im- proved collaboration and decision-making,”Applied Sciences, vol. 15, no. 7, p. 3794, 2025

    2025

  27. [32]

    Graph neural network-based multi-agent reinforcement learning for resilient distributed coordination of multi-robot systems,

    A. Goeckner, Y . Sui, N. Martinet, X. Li, and Q. Zhu, “Graph neural network-based multi-agent reinforcement learning for resilient distributed coordination of multi-robot systems,” in2024 IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS). IEEE, 2024, p...

    2024

  28. [33]

    Graph convolutional rein- forcement learning,

    J. Jiang, C. Dun, T. Huang, and Z. Lu, “Graph convolutional rein- forcement learning,”arXiv preprint arXiv:1810.09202, 2018

    2018 arXiv

  29. [34]

    Multi-agent graph- attention communication and teaming

    Y . Niu, R. R. Paleja, and M. C. Gombolay, “Multi-agent graph- attention communication and teaming.” inAAMAS, vol. 21, 2021, p. 20th

    2021

  30. [35]

    Leveraging graph neural networks and multi-agent reinforcement learning for inventory control in supply chains,

    N. Kotecha and A. del Rio Chanona, “Leveraging graph neural networks and multi-agent reinforcement learning for inventory control in supply chains,”Computers & Chemical Engineering, p. 109111, 2025

    2025

  31. [36]

    Geometric deep learning: Grids, groups, graphs, geodesics, and gauges,

    M. M. Bronstein, J. Bruna, T. Cohen, and P. Veli ˇckovi´c, “Geometric deep learning: Grids, groups, graphs, geodesics, and gauges,”arXiv preprint arXiv:2104.13478, 2021

    2021 arXiv

  32. [37]

    Learning color equivariant representations,

    Y . Yang, F. O’Mahony, and C. Allen-Blanchette, “Learning color equivariant representations,”arXiv preprint arXiv:2406.09588, 2024

    2024 arXiv

  33. [38]

    Gagrasp: Geometric algebra diffusion for dexterous grasping,

    T. Zhong and C. Allen-Blanchette, “Gagrasp: Geometric algebra diffusion for dexterous grasping,”arXiv preprint arXiv:2503.04123, 2025

    2025 arXiv

  34. [40]

    Boosting multiagent reinforcement learning via permutation invariant and permutation equivariant networks,

    H. Jianye, X. Hao, H. Mao, W. Wang, Y . Yang, D. Li, Y . Zheng, and Z. Wang, “Boosting multiagent reinforcement learning via permutation invariant and permutation equivariant networks,” inThe eleventh international conference on learning representations, 2022

    2022

  35. [41]

    Se (3)-equivariant robot learning and con- trol: A tutorial survey,

    J. Seo, S. Yoo, J. Chang, H. An, H. Ryu, S. Lee, A. Kruthiventy, J. Choi, and R. Horowitz, “Se (3)-equivariant robot learning and con- trol: A tutorial survey,”International Journal of Control, Automation and Systems, vol. 23, no. 5, pp. 1271–1306, 2025

    2025

  36. [43]

    Group equivariant convolutional networks,

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

    2016

  37. [44]

    Spherical cnns,

    T. S. Cohen, M. Geiger, J. Köhler, and M. Welling, “Spherical cnns,” arXiv preprint arXiv:1801.10130, 2018

    2018 arXiv

  38. [45]

    Steerable cnns,

    T. S. Cohen and M. Welling, “Steerable cnns,”arXiv preprint arXiv:1612.08498, 2016

    2016 arXiv

  39. [46]

    Learning so (3) equivariant representations with spherical cnns,

    C. Esteves, C. Allen-Blanchette, A. Makadia, and K. Daniilidis, “Learning so (3) equivariant representations with spherical cnns,” in Proceedings of the european conference on computer vision (ECCV), 2018, pp. 52–68

    2018

  40. [47]

    Equivariant multi-view networks,

    C. Esteves, Y . Xu, C. Allen-Blanchette, and K. Daniilidis, “Equivariant multi-view networks,” inProceedings of the IEEE/CVF international conference on computer vision, 2019, pp. 1568–1577

    2019

  41. [48]

    Sensing flow gradients is necessary for learning autonomous underwater navigation,

    Y . Jiao, H. Hang, J. Merel, and E. Kanso, “Sensing flow gradients is necessary for learning autonomous underwater navigation,”Nature Communications, vol. 16, no. 1, p. 3044, 2025

    2025

  42. [49]

    Do transformers really perform badly for graph representation?

    C. Ying, T. Cai, S. Luo, S. Zheng, G. Ke, D. He, Y . Shen, and T.-Y . Liu, “Do transformers really perform badly for graph representation?” Advances in neural information processing systems, vol. 34, pp. 28 877–28 888, 2021

    2021

  43. [50]

    Benchmarking graphormer on large-scale molecular modeling datasets,

    Y . Shi, S. Zheng, G. Ke, Y . Shen, J. You, J. He, S. Luo, C. Liu, D. He, and T.-Y . Liu, “Benchmarking graphormer on large-scale molecular modeling datasets,”arXiv preprint arXiv:2203.04810, 2022

    2022 arXiv

  44. [52]

    Semi-supervised classification with graph convolutional networks,

    T. Kipf, “Semi-supervised classification with graph convolutional networks,”arXiv preprint arXiv:1609.02907, 2016

    2016 arXiv

  45. [53]

    Graph attention networks,

    P. Veli ˇckovi´c, G. Cucurull, A. Casanova, A. Romero, P. Lio, and Y . Bengio, “Graph attention networks,”arXiv preprint arXiv:1710.10903, 2017

    2017 arXiv

  46. [54]

    Attention is all you need,

    A. Vaswani, N. Shazeer, N. Parmar, J. Uszkoreit, L. Jones, A. N. Gomez, Ł. Kaiser, and I. Polosukhin, “Attention is all you need,” Advances in neural information processing systems, vol. 30, 2017

    2017

  47. [55]

    Emergence of grounded com- positional language in multi-agent populations,

    I. Mordatch and P. Abbeel, “Emergence of grounded com- positional language in multi-agent populations,”arXiv preprint arXiv:1703.04908, 2017

    2017 arXiv

  48. [56]

    Pettingzoo: Gym for multi-agent reinforcement learning,

    J. Terry, B. Black, N. Grammel, M. Jayakumar, A. Hari, R. Sullivan, L. S. Santos, C. Dieffendahl, C. Horsch, R. Perez-Vicenteet al., “Pettingzoo: Gym for multi-agent reinforcement learning,”Advances in Neural Information Processing Systems, vol. 34, pp. 15 032–15 043, 2021

    2021

  49. [57]

    Curriculum learning,

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

    2009

  50. [58]

    Teacher–student curriculum learning,

    T. Matiisen, A. Oliver, T. Cohen, and J. Schulman, “Teacher–student curriculum learning,”IEEE transactions on neural networks and learning systems, vol. 31, no. 9, pp. 3732–3740, 2019

    2019

  51. [59]

    An introduction to the kalman filter,

    G. Welch, G. Bishopet al., “An introduction to the kalman filter,” 1995

    1995

  52. [60]

    Be- yond canonicalization: How tensorial messages improve equivariant message passing,

    P. Lippmann, G. Gerhartz, R. Remme, and F. A. Hamprecht, “Be- yond canonicalization: How tensorial messages improve equivariant message passing,”arXiv preprint arXiv:2405.15389, 2024

    2024 arXiv

Pith tools

Reviewed August 15, 2026 · model on record in the stance chip above.