Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation

Chengyang He; Gadiel Sznaier Camps; Guillaume Sartoretti; Mac Schwager; Xu Liu

REVIEW 5 major objections 5 minor 1 cited by

Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation

T0 review · 5 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 The paper claims that a decentralized diffusion policy can match a centralized controller on bimanual manipulation by giving each arm a private ego embedding and a shared consensus embedding.

desk verdict A plausible decentralized diffusion architecture for bimanual manipulation, but the empirical support for the headline no-communication mode is weak and the strong results come from the communication-enabled variant. read the letter →

arxiv 2505.09144 v1 pith:MKNH2L4F submitted 2025-05-14 cs.RO

Chengyang He , Gadiel Sznaier Camps , Xu Liu , Mac Schwager , Guillaume Sartoretti This is my paper · ORCID

classification cs.RO

keywords decentralizeddiffusionpolicycooperativemanipulationconsensuslearningsheaftheoryofmindbimanualmulti-armrobotsLaplacian

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 tries to establish that decentralized robot arms need not sacrifice coordination: each arm can run its own diffusion policy while coordinating through a latent consensus embedding trained to be common across agents. The authors claim that a sheaf-theoretic alignment loss, together with theory-of-mind and directional-consensus constraints, keeps this consensus embedding both consistent and informative. In hardware trials on two bimanual tasks, the resulting policy outperforms naive decentralized diffusion baselines and reaches performance comparable to a centralized diffusion policy, with or without a single communication step. If true, this would remove a central obstacle to scaling multi-arm manipulation without a fused central controller.

What carries the argument

The central object is the consensus embedding $h^{con}_u$: a latent vector each agent produces from shared third-person observations, trained so that the two agents' vectors agree despite different poses and private sensor streams. The argument is carried by three training losses and one inference operator: the first-order cohomology loss $L_{nc}$ enforces numerical agreement; the theory-of-mind loss $L_{tom}$ requires each agent to predict the other's ego embedding from its own consensus embedding, preserving information; the confidence loss $L_{conf}$ gives asymmetric alignment guided by per-agent confidence; and at inference the sheaf Laplacian update $h_{u,t+1} = (1-\eta) h_{u,t} + \eta h_{v,t}$ is a one-step consensus repair. Together these define a 'global section' of a cellular sheaf over the robot graph, in which each arm's latent interpretation of the shared scene is consistent with the other's.

What would settle it

Measure the per-step distance between the two arms' consensus embeddings in no-communication rollouts: if failures consistently coincide with the distance growing beyond a small threshold, the synchronization premise is the active failure mode. A complementary test would cover the paper's stated limits by adding a third arm or occluding the third-person camera; if performance drops to the naive decentralized baseline, the centralized-level claim depends on the shared camera view rather than on the consensus mechanism.

Watch

Extended reading notes

Core claim

On the paper's own terms, the discovery is that a decentralized multi-arm diffusion policy can achieve centralized-level coordination by splitting each agent's latent state into a private ego embedding and a shared consensus embedding. The consensus encoders are supervised during centralized training by a first-order sheaf cohomology loss that pushes the two arms' consensus embeddings toward global consistency, while a theory-of-mind decoder forces the consensus embedding to retain enough information to infer the other arm's ego state, preventing collapse. A directional consensus mechanism then lets the lower-confidence embedding align to the higher-confidence one. At inference, the arms can act with no explicit communication, or they can exchange consensus embeddings once per inference step and repair drift with a sheaf Laplacian consistency operator; the communicated variant is the one that matches the centralized baseline most reliably.

Load-bearing premise

The load-bearing premise is that the two arms' independently computed consensus embeddings stay close enough during inference to keep their decentralized policies coordinated without explicit communication; the paper itself concedes in Section 3.2 that it cannot guarantee this, which is why its best results use a one-step sheaf Laplacian exchange, and its Limitations section adds that only two arms and a fixed third-person camera were tested.

Editorial extensions

If this is right

If the claim holds, a team of manipulators can be trained with a single fused dataset and deployed with each arm running its own policy, so a failed or delayed arm does not crash the whole controller.
The one-step sheaf Laplacian exchange gives a cheap, model-agnostic way to correct consensus drift at inference, converting a decentralized policy's main weakness into a small communication cost.
The theory-of-mind constraint suggests consensus latents can double as an intent signal: each arm's consensus embedding directly supports predicting the other arm's private state.
The architecture is stated to generalize to any number of agents on a graph, so the same losses would apply to multi-arm systems beyond two robots.

Reading between the lines

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

Editorial inference: the no-communication variant's reliability should degrade as the two arms' shared third-person observations diverge, and measuring consensus-embedding distance during rollouts would test whether drift predicts failure.
Editorial inference: the learned confidence scores could serve as a perceptual-quality signal, letting a robot decide when to trust its own consensus embedding versus defer to a teammate.
Editorial inference: the same consensus-plus-ego split could apply to human-robot collaboration, with the theory-of-mind predictor modeling a human's intent from shared scene context, although the paper does not test this.
Editorial inference: because the paper reports results only for two arms and a fixed third-person camera, its scalability claim rests on untested generalization; adding a third arm or occluding the shared view would be a natural stress test.

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.

Desk Editor's Note

read the letter

The core idea here is worth taking seriously: separate each arm's representation into an ego embedding and a consensus embedding, use a sheaf-inspired consistency loss during centralized training, and add a theory-of-mind decoder plus a confidence-weighted alignment so the consensus stays expressive. That combination is not in the cited literature, and the hardware evaluation on two real bimanual tasks is a genuine step beyond simulation-only work. The sheaf-Laplacian repair as an optional one-step communication mechanism is a clean practical trick, and the paper is honest that it requires a message exchange.

The soft spots are real but not fatal to the architecture. The statistics are thin: 15 rollouts per method, no seeds, no error bars, no significance tests. Task 1 is qualitative only. The paper's own Section 3.2 admits that consensus embeddings cannot be guaranteed to stay synchronized without communication, and the data confirm it: LatentToM without communication gets 13/15 and a "partially complete" on Task 1, whereas the sheaf-Laplacian version gets 14/15 and "fully complete." So the headline claim of matching a centralized policy really rests on the communicating variant. Also, both arms are fed the same fixed third-person camera image for the consensus input, which weakens the "different sensor streams" story—the consensus is anchored by a shared global observation, not by genuinely distinct local views. The directional consensus loss in Eq. (4) is written as a symmetric distance scaled by the confidence difference plus an entropy term; as stated, it does not literally implement one-way alignment, and the paper should either fix the equation or clarify the mechanism. The robustness-to-failure claim is asserted in the abstract but never tested, and the paper says code will be released but gives no link or hyperparameters.

Who is this for? Anyone working on decentralized multi-arm manipulation or on making diffusion policies multi-agent. The architecture is a reasonable starting point, and the sheaf-consistency framing is a useful way to think about shared latent spaces, even if the math here is mostly a regularizer rather than a deep theoretical result. I would not desk-reject it. It deserves a serious referee, but the referee should push for more trials, error bars, a test of the failure/delay scenario, a corrected or clarified Eq. (4), and actual code. The central idea is not disproven by these weaknesses; it is just under-supported.

Referee Report

5 major / 5 minor

Summary. The paper proposes LatentToM, a decentralized diffusion policy architecture for bimanual cooperative manipulation. Each arm maintains an ego embedding and a consensus embedding; training uses a sheaf-inspired consistency loss, a theory-of-mind loss that predicts the other arm's ego embedding, and a confidence-weighted directional consensus loss. At inference, execution can be fully distributed without explicit communication, or it can use a one-step sheaf-Laplacian exchange to align consensus embeddings. Hardware experiments on a Push-T task and a coffee-pouring task compare LatentToM with a centralized diffusion policy and decentralized baselines, and the paper claims that LatentToM outperforms the naive decentralized baseline and is comparable to the centralized policy. The manuscript also includes a limitations section noting that only two-arm systems were tested and that the method relies on a fixed third-person camera.

Significance. If the central claims held, the architecture would be a useful step toward decentralized multi-arm manipulation, particularly the no-communication mode that maintains coordination from local observations alone. The paper has concrete strengths: real hardware rollouts on two tasks, a fine-grained outcome taxonomy in Table 1, an explicit if partial limitations section, and a clean proof in Appendix B that Eq. (5) is a Laplacian step under the stated assumptions. However, the contribution is not yet established. The theory-of-mind loss as written is trivially satisfiable because the ground-truth ego embedding is an input to the predictor, the directional consensus loss is symmetric rather than directional, and the empirical support for the no-communication mode is statistically underpowered and confounded by the shared third-person camera. Under the paper's own assumptions, the sheaf-theoretic construction reduces to ordinary pairwise L2 consistency. These issues bear directly on the paper's stated claims, so the result should be considered promising but not yet proven.

major comments (5)

[Eq. (3), Appendix A.1] Equation (3) defines the ToM loss with the ground-truth ego embedding h_ego_v as an input to the predictor ψ_{u→v}; a predictor that ignores h_con_u and copies its second argument drives the loss to zero, so L_tom does not force the consensus embedding to carry information about the other agent. Appendix A.1 confirms this reading by making h_ego_v the key and value in the attention module, and this setup creates a train/test mismatch because h_ego_v is not available at decentralized inference. The loss must be reformulated without the target as an input, and the experiments rerun, before the claims about the ToM constraint can be evaluated.
[Eq. (4)] The directional consensus loss is symmetric: because ||h_v−h_u|| equals ||h_u−h_v||, the two indicator-weighted terms are identical and Eq. (4) reduces to (1+|Δc|)||h_u−h_v|| + λ_ent(H(c_u)+H(c_v)). Confidence scores therefore only scale a symmetric distance; the objective does not implement one-way alignment of the lower-confidence embedding toward the higher-confidence one, contrary to the description in Section 3.1. Please either replace this loss with a genuinely asymmetric objective (e.g., a stop-gradient or one-sided norm) or revise the claims.
[Section 4.2.2, Table 1] The quantitative support for the headline claim is 15 rollouts per method from a single run. The reported differences (CDP 15/15, LatentToM with SL 14/15, LatentToM without communication 13/15) are not accompanied by seeds, variance, confidence intervals, or significance tests, so 'comparable performance' is not statistically established. The no-communication mode—the mode that would make the decentralized claim nontrivial—has no quantitative evaluation on Task 1 (only 'partially complete' vs 'fully complete'), and Section 3.2 states that consensus-embedding synchronization cannot be guaranteed without the sheaf-Laplacian communication step. The strongest results therefore come from the variant that communicates.
[Section 3.1, Section 6] Both arms receive the same fixed third-person camera image as o_con, so Eq. (2) aligns outputs of two consensus encoders on identical input. This confounds the 'without explicit communication' comparison: the consensus embedding is anchored by a common global observation rather than by genuinely distinct local sensor streams. Section 6 acknowledges the reliance on the fixed third-person camera, but the paper should either ablate this factor (e.g., per-agent views or occlusions) or substantially temper the claim that coordination emerges from independent perception.
[Eqs. (1)-(2), Appendix B] The sheaf-theoretic construction is a special case: the node spaces and edge space are all R^dcon and the restriction maps are identity. Under these assumptions Eq. (2) is ordinary pairwise L2 consistency and Eq. (5) is the standard two-node graph Laplacian average. The Appendix B proof is correct for this special case, but it does not justify describing the method as a general 'first-order cohomology loss' or 'sheaf consistency' mechanism. The paper should either use nontrivial restriction maps or edge spaces, or present the contribution as a consensus regularizer with auxiliary structural losses.

minor comments (5)

[Section 4.2.2] Report the number of random seeds and per-seed results, and provide standard errors or confidence intervals for the success rates; 15 rollouts without variance is insufficient to distinguish 14/15 from 15/15.
[Section 3.1] The hyperparameters α, β, γ, λ_ent, and η are introduced but no values, ranges, or sensitivity analysis are given despite the statement that they 'must be carefully tuned.' Add an implementation appendix with the chosen values and at least a small ablation.
[Abstract, Section 4.2] Calling the vanilla Diffusion Policy a 'state-of-the-art centralized diffusion policy' overstates the baseline, since no comparison is made to other centralized multi-agent diffusion methods such as MADiff. Please temper the wording or broaden the comparison.
[Appendix A.2, Figure 5] Appendix A.2 contains the phrase 'The Specifically,' which appears to be a typo, and Section 4.2.2 uses 'NNDP' while the method is introduced as 'NDDP.' Please standardize the terminology.
[Conclusion, Section 6] The conclusion asserts that the approach 'can naturally be extended to an arbitrary number of agents' without evidence beyond two-arm experiments; Section 6's own limitations should be reflected in the abstract and conclusion wording.

Circularity Check

1 steps flagged · score 6.0 of 10

The paper's hardware comparison is not circular, but the ToM 'prediction' loss reduces by construction because the target ego embedding is fed into the predictor.

self definitional [Section 3.1, Eq. (3); Appendix A.1 (ToM Predictor)]
"'we introduce an additional loss function Ltom: Ltom = P (u,v)∈E ||hego v −ψu→v(hcon u ,h ego v )||2 2 + (v↔u)' and 'the consensus embedding of the ego arm hcon u serves as the query, while the ego embedding of v acts as both the key and value.'"

The loss is meant to force each arm to infer the other's ego embedding from its own consensus embedding, but the target hego v is itself an input to the predictor ψu→v. With the attention implementation using hego v as key and value, the network can drive Ltom to zero by copying the target from its input, so the loss does not actually constrain the consensus embedding to carry the other agent's state. The claimed ToM structural constraint is therefore satisfiable by construction regardless of the consensus content, making the 'expressiveness' justification circular rather than a derivation.

full rationale

The paper's main empirical claim—that LatentToM matches a centralized diffusion policy on hardware—is self-contained in the sense that it compares rollout outcomes on external tasks; no success metric is a function of a fitted parameter, and the sheaf-Laplacian claim in Appendix B is an explicit derivation from identity restriction maps. The self-citation [17] is not load-bearing: Eq. (2) is a plain pairwise agreement loss whose effect is apparent without the citation. However, one load-bearing internal step is circular: the ToM loss in Eq. (3) and its attention implementation in Appendix A.1 feed the target ego embedding hego v into the predictor as key/value, so minimizing Ltom can be achieved by copying the input, not by reading the consensus embedding. That makes the stated mechanism by which the consensus embedding is kept expressive (the ToM constraint) self-definitional. The no-communication and small-rollout concerns are correctness/statistical risks, not circularity, and do not affect this score beyond the one identified step.

Assumptions & free parameters 6 free parameters · 5 assumptions · 0 invented entities

No new physical entities are postulated. The consensus embedding, ego embedding, confidence score, and ToM predictor are learned representations inside the policy, not independent entities with falsifiable handles. The central claim rests mainly on an empirical comparison, with the sheaf-theoretic component acting as a regularizer and a simple Laplacian repair step.

free parameters (6)

alpha (weight of L_nc)
Total auxiliary loss weight in L_tot = alpha*L_nc + beta*L_tom + gamma*L_conf; Section 3.1 says these 'must be carefully tuned' and no values are reported.
beta (weight of L_tom)
Weight for the theory-of-mind loss; tuned per task and not reported.
gamma (weight of L_conf)
Weight for the confidence loss; tuned per task and not reported.
lambda_ent (entropy regularization in Eq. 4)
Coefficient for the confidence entropy term; value not reported.
eta (sheaf Laplacian step size)
Step size in Eq. (5) for online consensus adjustment; value not reported and likely task-dependent.
dimensions d_con and d_ego
Dimensionality of consensus and ego embeddings chosen by hand; no ablation on capacity is given.

assumptions (5)

standard math Standard sheaf theory definitions (cellular sheaf, restriction maps, cohomology, Laplacian) from Bredon [43] and Hansen/Ghrist [44] are used without formal justification.
The paper relies on this background to motivate L_nc and the sheaf Laplacian update.
ad hoc to paper Restriction maps are the identity and the edge space equals R^d_con.
Section 3.1 sets F(e)=R^d_con and Appendix B assumes rho=I, which reduces the sheaf machinery to a graph Laplacian. This simplification is load-bearing for the proof but not derived from task structure.
domain assumption Both agents receive the same third-person camera observation o_con.
Consensus alignment in the no-communication mode relies on this shared input; Section 3.1 defines o_con as the third-person view. Without it, the consensus embeddings have no common grounding.
domain assumption Centralized demonstrations from a single user controlling both arms are sufficient to train decentralized policies.
Section 2 describes data collection for centralized policies, but the paper does not explain how the demos are assigned to each arm's decentralized policy or how consensus labels arise.
domain assumption Adding auxiliary losses to the diffusion objective does not destabilize training.
The paper adds L_nc, L_tom, and L_conf to the standard diffusion training but provides no stability analysis or learning-curve evidence.

how reviews work

0 comments

Cite this review

Pith. "Pith review of Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation." pith.science (2026). https://pith.science/paper/MKNH2L4F

@misc{pith2026250509144,
  author       = {Pith},
  title        = {Pith review of: Latent Theory of Mind: A Decentralized Diffusion Architecture for Cooperative Manipulation},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/MKNH2L4F}},
  note         = {Machine review of arXiv:2505.09144}
}

read the original abstract

We present Latent Theory of Mind (LatentToM), a decentralized diffusion policy architecture for collaborative robot manipulation. Our policy allows multiple manipulators with their own perception and computation to collaborate with each other towards a common task goal with or without explicit communication. Our key innovation lies in allowing each agent to maintain two latent representations: an ego embedding specific to the robot, and a consensus embedding trained to be common to both robots, despite their different sensor streams and poses. We further let each robot train a decoder to infer the other robot's ego embedding from their consensus embedding, akin to theory of mind in latent space. Training occurs centrally, with all the policies' consensus encoders supervised by a loss inspired by sheaf theory, a mathematical theory for clustering data on a topological manifold. Specifically, we introduce a first-order cohomology loss to enforce sheaf-consistent alignment of the consensus embeddings. To preserve the expressiveness of the consensus embedding, we further propose structural constraints based on theory of mind and a directional consensus mechanism. Execution can be fully distributed, requiring no explicit communication between policies. In which case, the information is exchanged implicitly through each robot's sensor stream by observing the actions of the other robots and their effects on the scene. Alternatively, execution can leverage direct communication to share the robots' consensus embeddings, where the embeddings are shared once during each inference step and are aligned using the sheaf Laplacian. In our hardware experiments, LatentToM outperforms a naive decentralized diffusion baseline, and shows comparable performance with a state-of-the-art centralized diffusion policy for bi-manual manipulation. Project website: https://stanfordmsl.github.io/LatentToM/.

Figures

Figures reproduced from arXiv: 2505.09144 by the authors.

**Figure 1.** Multi-arm robotic system. In our setup, the system consists of two robotic arms, each equipped with an end-effector camera, represented by the red and green areas indicating their respective fields of view. Additionally, a third-person camera observes the overlapping workspace between the two arms, shown in gray. The bottom part illustrates the consensus embeddings generated using sheaf theory from our collected … view at source ↗

**Figure 2.** Cooperative Manipulation Tasks. In (a), the yellow dots and lines represent the ideal [PITH_FULL_IMAGE:figures/full_fig_p006_2.png] view at source ↗

**Figure 3.** T block in the fully InD (3a,3b) and partially OOD (asymmetric friction coefficients, 3c) cases. In this section, we conduct comparative experiments using five methods across Tasks 1 and 2: (1) Centralized Diffusion Policy (i.e., vanilla diffusion policy); (2) Naive Decentralized Diffusion Policy (where we only use its own end-effector and third-view cameras during training for each arm’s policy); (3) Naive Consen… view at source ↗

Figures from the paper (3 more)

**Figure 4.** Figure 4: Rollout results of different methods under unbalanced friction setup as shown in Fig [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]

**Figure 5.** Figure 5: Representative rollout results for Task 2 (coffee bean pouring). As shown in the figure, [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]

**Figure 6.** Figure 6: ToM Predictor. As illustrated in [PITH_FULL_IMAGE:figures/full_fig_p012_6.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

Reviewed papers in the Pith corpus that reference this work. Sorted by Pith novelty score. Full citation record

Think Proprioceptively: State-Grounded Visual Token Selection for VLA Policies
cs.RO 2026-02 conditional novelty 6.0 of 10

Using tokenized proprioception plus instruction to select ~15% of visual patches matches or beats full-token VLA baselines and cuts latency by ~58%.

Reference graph

Works this paper leans on

45 extracted references · 19 canonical work pages · cited by 1 Pith paper

[1]

Billard and D

A. Billard and D. Kragic. Trends and challenges in robot manipulation. Science, 364(6446): eaat8414, 2019

2019
[2]

D ¨omel, S

A. D ¨omel, S. Kriegel, M. Kaßecker, M. Brucker, T. Bodenm ¨uller, and M. Suppa. Toward fully autonomous mobile manipulation for industrial environments. International Journal of Advanced Robotic Systems, 14(4):1729881417718588, 2017

work page 2017
[3]

Benali, J.-F

K. Benali, J.-F. Breth ´e, F. Gu ´erin, and M. Gorka. Dual arm robot manipulator for grasping boxes of different dimensions in a logistics warehouse. In2018 IEEE International Conference on Industrial Technology (ICIT), pages 147–152. IEEE, 2018

work page 2018
[4]

Ginoya, Y

T. Ginoya, Y . Maddahi, and K. Zareinia. A historical review of medical robotic platforms. Journal of Robotics, 2021(1):6640031, 2021

work page 2021
[5]

D. Guri, M. Lee, O. Kroemer, and G. Kantor. Hefty: A modular reconfigurable robot for advancing robot manipulation in agriculture. arXiv preprint arXiv:2402.18710, 2024

arXiv 2024
[6]

Y . Ze, Z. Chen, W. Wang, T. Chen, X. He, Y . Yuan, X. B. Peng, and J. Wu. Generalizable humanoid manipulation with improved 3d diffusion policies.arXiv preprint arXiv:2410.10803, 2024

arXiv 2024
[7]

F. AI. Helix: A vision-language-action model for generalist humanoid control. https:// www.figure.ai/news/helix, 2025. Accessed: 2025-02-20

work page 2025
[8]

C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song. Diffusion policy: Visuomotor policy learning via action diffusion. The International Journal of Robotics Research, page 02783649241273668, 2023

2023

Show all 45 references

[9]

N. M. Shafiullah, Z. Cui, A. A. Altanzaya, and L. Pinto. Behavior transformers: Cloning k modes with one stone. Advances in neural information processing systems, 35:22955–22968, 2022

2022
[10]

Mandlekar, D

A. Mandlekar, D. Xu, J. Wong, S. Nasiriany, C. Wang, R. Kulkarni, L. Fei-Fei, S. Savarese, Y . Zhu, and R. Mart´ın-Mart´ın. What matters in learning from offline human demonstrations for robot manipulation. arXiv preprint arXiv:2108.03298, 2021

2021 arXiv
[11]

Florence, C

P. Florence, C. Lynch, A. Zeng, O. A. Ramirez, A. Wahid, L. Downs, A. Wong, J. Lee, I. Mor- datch, and J. Tompson. Implicit behavioral cloning. In Conference on robot learning, pages 158–168. PMLR, 2022. 9

2022
[12]

Gupta, V

A. Gupta, V . Kumar, C. Lynch, S. Levine, and K. Hausman. Relay policy learning: Solving long-horizon tasks via imitation and reinforcement learning.arXiv preprint arXiv:1910.11956, 2019

1910 arXiv
[13]

J. Ho, A. Jain, and P. Abbeel. Denoising diffusion probabilistic models. Advances in neural information processing systems, 33:6840–6851, 2020

2020
[14]

H. Ha, J. Xu, and S. Song. Learning a decentralized multi-arm motion planner. arXiv preprint arXiv:2011.02608, 2020

2011 arXiv
[15]

Sartoretti, J

G. Sartoretti, J. Kerr, Y . Shi, G. Wagner, T. S. Kumar, S. Koenig, and H. Choset. Primal: Pathfinding via reinforcement and imitation multi-agent learning. IEEE Robotics and Automa- tion Letters, 4(3):2378–2385, 2019

2019
[16]

C. He, T. Duhan, P. Tulsyan, P. Kim, and G. Sartoretti. Social behavior as a key to learning- based multi-agent pathfinding dilemmas. arXiv preprint arXiv:2408.03063, 2024

2024 arXiv
[17]

S. Liao, W. Xia, Y . Cao, W. Dai, C. He, W. Wu, and G. Sartoretti. Sigma: Sheaf-informed geometric multi-agent pathfinding. arXiv preprint arXiv:2502.06440, 2025

2025
[18]

Z. Zhu, H. Zhao, H. He, Y . Zhong, S. Zhang, H. Guo, T. Chen, and W. Zhang. Diffusion models for reinforcement learning: A survey. arXiv preprint arXiv:2311.01223, 2023

2023 arXiv
[19]

Y . Ze, G. Zhang, K. Zhang, C. Hu, M. Wang, and H. Xu. 3d diffusion policy. arXiv e-prints, pages arXiv–2403, 2024

2024
[20]

D. Wang, S. Hart, D. Surovik, T. Kelestemur, H. Huang, H. Zhao, M. Yeatman, J. Wang, R. Walters, and R. Platt. Equivariant diffusion policy.arXiv preprint arXiv:2407.01812, 2024

2024 arXiv
[21]

Yang, Z.-a

J. Yang, Z.-a. Cao, C. Deng, R. Antonova, S. Song, and J. Bohg. Equibot: Sim (3)-equivariant diffusion policy for generalizable and data efficient learning.arXiv preprint arXiv:2407.01479, 2024

2024 arXiv
[22]

U. A. Mishra, S. Xue, Y . Chen, and D. Xu. Generative skill chaining: Long-horizon skill planning with diffusion models. In Conference on Robot Learning, pages 2905–2925. PMLR, 2023

2023
[23]

A. Z. Ren, J. Lidard, L. L. Ankile, A. Simeonov, P. Agrawal, A. Majumdar, B. Burch- fiel, H. Dai, and M. Simchowitz. Diffusion policy policy optimization. arXiv preprint arXiv:2409.00588, 2024

2024 arXiv
[24]

Y . Wang, Y . Zhang, M. Huo, R. Tian, X. Zhang, Y . Xie, C. Xu, P. Ji, W. Zhan, M. Ding, et al. Sparse diffusion policy: A sparse, reusable, and flexible policy for robot learning. arXiv preprint arXiv:2407.01531, 2024

2024 arXiv
[25]

X. Ma, S. Patidar, I. Haughton, and S. James. Hierarchical diffusion policy for kinematics- aware multi-task robotic manipulation. In Proceedings of the IEEE/CVF Conference on Com- puter Vision and Pattern Recognition, pages 18081–18090, 2024

2024
[26]

Razmjoo, S

A. Razmjoo, S. Calinon, M. Gienger, and F. Zhang. Ccdp: Composition of conditional diffu- sion policies with guided sampling. arXiv preprint arXiv:2503.15386, 2025

2025
[27]

S. Fan, Q. Yang, Y . Liu, K. Wu, Z. Che, Q. Liu, and M. Wan. Diffusion trajectory-guided policy for long-horizon robot manipulation. arXiv preprint arXiv:2502.10040, 2025

2025
[28]

Y . Wu, Z. Chen, F. Wu, L. Chen, L. Zhang, Z. Bing, A. Swikir, S. Haddadin, and A. Knoll. Tacdiffusion: Force-domain diffusion policy for precise tactile manipulation. arXiv preprint arXiv:2409.11047, 2024. 10

2024 arXiv
[29]

K. Rana, J. Haviland, S. Garg, J. Abou-Chakra, I. Reid, and N. Suenderhauf. Sayplan: Ground- ing large language models using 3d scene graphs for scalable robot task planning. arXiv preprint arXiv:2307.06135, 2023

2023 arXiv
[30]

T.-W. Ke, N. Gkanatsios, and K. Fragkiadaki. 3d diffuser actor: Policy diffusion with 3d scene representations. arXiv preprint arXiv:2402.10885, 2024

2024 arXiv
[31]

Z. Fu, T. Z. Zhao, and C. Finn. Mobile aloha: Learning bimanual mobile manipulation with low-cost whole-body teleoperation. arXiv preprint arXiv:2401.02117, 2024

2024 arXiv
[32]

C. Chi, Z. Xu, C. Pan, E. Cousineau, B. Burchfiel, S. Feng, R. Tedrake, and S. Song. Universal manipulation interface: In-the-wild robot teaching without in-the-wild robots. arXiv preprint arXiv:2402.10329, 2024

2024 arXiv
[33]

Oroojlooy and D

A. Oroojlooy and D. Hajinezhad. A review of cooperative multi-agent deep reinforcement learning. Applied Intelligence, 53(11):13677–13722, 2023

2023
[34]

X. Lian, C. Zhang, H. Zhang, C.-J. Hsieh, W. Zhang, and J. Liu. Can decentralized algorithms outperform centralized algorithms? a case study for decentralized parallel stochastic gradient descent. Advances in neural information processing systems, 30, 2017

2017
[35]

Shome, K

R. Shome, K. Solovey, J. Yu, K. Bekris, and D. Halperin. Fast, high-quality dual-arm re- arrangement in synchronous, monotone tabletop setups. In International Workshop on the Algorithmic Foundations of Robotics, pages 778–795. Springer, 2018

2018
[36]

J. P. Van Den Berg and M. H. Overmars. Roadmap-based motion planning in dynamic envi- ronments. IEEE transactions on robotics, 21(5):885–897, 2005

2005
[37]

J. Li, Z. Chen, D. Harabor, P. J. Stuckey, and S. Koenig. Mapf-lns2: Fast repairing for multi- agent path finding via large neighborhood search. In Proceedings of the AAAI Conference on Artificial Intelligence, volume 36, pages 10256–10265, 2022

2022
[38]

Z. Zhu, M. Liu, L. Mao, B. Kang, M. Xu, Y . Yu, S. Ermon, and W. Zhang. Madiff: Of- fline multi-agent learning with diffusion models. Advances in Neural Information Processing Systems, 37:4177–4206, 2024

2024
[39]

Robinson

M. Robinson. Sheaves are the canonical data structure for sensor integration. Information Fusion, 36:208–224, 2017

2017
[40]

J. M. Curry. Sheaves, cosheaves and applications. University of Pennsylvania, 2014

2014
[41]

Robinson

M. Robinson. Understanding networks and their behaviors using sheaf theory. In 2013 IEEE Global Conference on Signal and Information Processing, pages 911–914. IEEE, 2013

2013
[42]

Bodnar, F

C. Bodnar, F. Di Giovanni, B. Chamberlain, P. Lio, and M. Bronstein. Neural sheaf diffusion: A topological perspective on heterophily and oversmoothing in gnns. Advances in Neural Information Processing Systems, 35:18527–18541, 2022

2022
[43]

G. E. Bredon. Sheaf theory, volume 170. Springer Science & Business Media, 2012

2012
[44]

Hansen and R

J. Hansen and R. Ghrist. Learning sheaf laplacians from smooth signals. InICASSP 2019-2019 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP) , pages 5446–5450. IEEE, 2019

2019
[45]

h(t+1) u h(t+1) v # = 1−η η η 1−η | {z } consistency operator ·

X. Wei and G.-W. Wei. Persistent sheaf laplacians. arXiv preprint arXiv:2112.10906, 2021. 11 A Implementation Details In this section, we provide detailed implementation details about the ToM Predictor and Confidence Predictor introduced in Section 3.1, where the main focus is...

2021 arXiv

Pith tools

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

[1] [1]

Billard and D

A. Billard and D. Kragic. Trends and challenges in robot manipulation. Science, 364(6446): eaat8414, 2019

2019

[2] [2]

D ¨omel, S

A. D ¨omel, S. Kriegel, M. Kaßecker, M. Brucker, T. Bodenm ¨uller, and M. Suppa. Toward fully autonomous mobile manipulation for industrial environments. International Journal of Advanced Robotic Systems, 14(4):1729881417718588, 2017

work page 2017

[3] [3]

Benali, J.-F

K. Benali, J.-F. Breth ´e, F. Gu ´erin, and M. Gorka. Dual arm robot manipulator for grasping boxes of different dimensions in a logistics warehouse. In2018 IEEE International Conference on Industrial Technology (ICIT), pages 147–152. IEEE, 2018

work page 2018

[4] [4]

Ginoya, Y

T. Ginoya, Y . Maddahi, and K. Zareinia. A historical review of medical robotic platforms. Journal of Robotics, 2021(1):6640031, 2021

work page 2021

[5] [5]

D. Guri, M. Lee, O. Kroemer, and G. Kantor. Hefty: A modular reconfigurable robot for advancing robot manipulation in agriculture. arXiv preprint arXiv:2402.18710, 2024

arXiv 2024

[6] [6]

Y . Ze, Z. Chen, W. Wang, T. Chen, X. He, Y . Yuan, X. B. Peng, and J. Wu. Generalizable humanoid manipulation with improved 3d diffusion policies.arXiv preprint arXiv:2410.10803, 2024

arXiv 2024

[7] [7]

F. AI. Helix: A vision-language-action model for generalist humanoid control. https:// www.figure.ai/news/helix, 2025. Accessed: 2025-02-20

work page 2025

[8] [8]

C. Chi, Z. Xu, S. Feng, E. Cousineau, Y . Du, B. Burchfiel, R. Tedrake, and S. Song. Diffusion policy: Visuomotor policy learning via action diffusion. The International Journal of Robotics Research, page 02783649241273668, 2023

2023