Pith. sign in

REVIEW 3 major objections 6 minor 31 references

CoDiMAD: Diffusion-Based Privileged Distillation for Communication-Free Multi-Robot Coordination

T0 review · 3 major / 6 minor · reviewed 2026-07-13 · grok-4.5

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

Pith's one-line read When robots only see local sensors, the same view can demand several valid joint moves; diffusion distillation samples those modes instead of averaging them into invalid actions.

desk verdict Solid compositional methods paper: privileged MAPPO → filtered offline pairs → conditional diffusion students, with a clean mode-averaging diagnosis that the experiments actually support. read the letter →

arxiv 2607.09587 v1 pith:73PV3ASR submitted 2026-07-10 cs.RO

Jiyue Tao , Shunheng Xin , Tongsheng Shen , Dexin Zhao , Feitian Zhang This is my paper · ORCID
classification cs.RO
keywords multi-robotsystemsprivilegedpolicydistillationpartialobservabilitydiffusionmulti-agentimitationlearningcommunication-freecoordinationMAPPODec-POMDP
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

This paper argues that communication-free multi-robot teams fail under partial observability because one local sensor reading can match several global situations that require qualitatively different cooperative actions. Privileged distillation first trains a fully informed oracle, then teaches sensor-limited students, but ordinary deterministic cloning collapses those options to their mean and produces hesitant or colliding behavior. CoDiMAD instead models the student as a conditional denoising diffusion policy that approximates the full oracle-action distribution and samples decisive actions from coherent modes. Theory shows why mean-squared regression lands in a low-density gap between modes, while reverse diffusion recovers the mixture. On coverage, pursuit-evasion, and box-pushing tasks with three agents, the method beats both direct local multi-agent RL and deterministic distillation while approaching the non-deployable oracle.

What carries the argument

CoDiMAD: a three-stage pipeline that trains a privileged MAPPO oracle, builds an offline dataset of local-observation–oracle-action pairs from successful collision-free rollouts, and distills into decentralized conditional DDPM students (accelerated by DDIM) that denoise actions given only local map history.

What would settle it

On the same three tasks, replace the diffusion student with an otherwise identical deterministic MSE head (or train diffusion on unfiltered oracle samples including failures) and check whether success rates and collisions reverse the gaps reported in Table II, especially the large box-pushing and collision improvements.

Watch

Extended reading notes

Core claim

Partial observability in multi-agent coordination induces a multi-modal conditional oracle-action distribution; deterministic regression-based distillation therefore converges to the conditional mean, which can lie between valid modes and yield invalid actions, whereas students parameterized as conditional diffusion models approximate that distribution and can sample from individual coordination modes.

Load-bearing premise

The student only ever sees filtered mean actions from successful, collision-free oracle episodes, so useful multi-modal coordination is assumed to be fully represented by that offline support and not to require recovery skills outside it.

Editorial extensions

If this is right

  • Communication-free robot teams can inherit most of a privileged oracle’s coordination skill without runtime message passing.
  • Deterministic behavioral cloning is systematically unsafe when local observations leave multiple joint strategies open.
  • Diffusion policies become a practical student class for multi-robot privileged distillation once few-step DDIM sampling is used.
  • Action-space and trajectory visualizations can diagnose mode averaging versus mode recovery in multi-agent imitation.

Reading between the lines

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

  • Any multi-agent setting whose local sensors hide teammate intent (not only marine robots) is a candidate for the same multi-modal failure mode.
  • If online fine-tuning with diffusion-compatible RL is added, the offline support limitation may shrink enough for physical deployment under distribution shift.
  • Attention-based denoisers suggested for larger teams would also test whether mode recovery scales when the number of latent global configurations grows combinatorially.
Share X Bluesky LinkedIn Reddit HN

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 6 minor

Summary. The paper proposes CoDiMAD, a three-stage privileged-distillation framework for communication-free multi-robot coordination under partial observability. A MAPPO oracle is trained with global state, an offline dataset of local-observation–oracle-action pairs is collected from successful collision-free rollouts (using the oracle mean action), and decentralized student policies are trained as conditional DDPMs (with DDIM inference) that act only on local BEV observations. The central claim is that partial observability induces multi-modal conditional oracle-action distributions (Proposition 1), that MSE-based deterministic distillation collapses to a low-density conditional mean between modes (Corollary 1), and that diffusion students recover coherent modes, yielding higher task success and fewer collisions than MAPPO-Local and BC-RNN on Coverage, Pursuit-Evasion, and Box Pushing (Table II), with supporting action- and trajectory-level visualizations (Figs. 3–4).

Significance. If the result holds, the paper makes a clear and useful contribution to multi-robot learning under communication constraints: it identifies mode averaging as a structural failure mode of standard privileged distillation in multi-agent settings, gives a clean formalization of that failure, and shows that conditional diffusion policies are an effective remedy without requiring inter-agent communication or human demonstrations. Strengths include a coherent three-stage pipeline, explicit theory–experiment linkage (Corollary 1 vs. Fig. 3 and BC-RNN collision spikes), controlled ablations (RNN, privileged upper bound, local MARL, deterministic BC), and multi-seed evaluation with task returns rather than training-loss proxies. The work is timely for marine and other bandwidth-limited multi-robot domains and is among the first to combine privileged multi-agent RL with diffusion-based student policies.

major comments (3)
  1. [§V-C, Table II; §IV-D] The central methodological claim is that multi-modality under partial observability requires modeling a full conditional action distribution rather than a point estimate. The only generative/multi-modal comparison is against deterministic MSE BC-RNN (Table II, §V-C). Without at least one alternative multi-modal student class on the same offline dataset (e.g., a Gaussian mixture / MDN head, CVAE, or simple energy-based policy), it remains unclear whether the gains are due to diffusion’s reverse process specifically or to any capacity to represent multiple modes. Adding one such baseline would substantially strengthen the claim that diffusion is the appropriate tool.
  2. [§IV-B, Algorithm 1; §VI] Stage 2 retains only successful, collision-free oracle episodes and labels them with the deterministic mean of the oracle Gaussian (Algorithm 1 lines 17–19; §IV-B). Multi-modality is then induced by different global configurations that share a local observation, which is valid for the reported BC-RNN vs. CoDiMAD comparison on the same support. However, absolute claims about recovering “coherent coordination modes” and approaching the oracle under decentralized execution rest on that filtered support covering the modes needed at test time. The authors acknowledge offline OOD fragility in §VI; a brief quantitative check (e.g., fraction of evaluation states near the dataset support, or a controlled recovery/perturbation experiment) would make the load-bearing assumption explicit rather than left to the limitation paragraph.
  3. [§V-C, Table II (Box Pushing)] On Box Pushing, CoDiMAD reaches 72.2% success versus the oracle’s 98.2% and BC-RNN’s 6.5% (Table II). This is the largest residual gap to the privileged upper bound and the setting where force coordination is tightest, yet the paper offers little failure-mode analysis (timing, contact geometry, relative pose ambiguity). Without that analysis it is hard to tell how much of the remaining gap is irreducible partial-observability ambiguity versus distillation or capacity limits, which weakens the interpretation of “approaching oracle-level performance” across tasks.
minor comments (6)
  1. [§IV-D; Fig. 3] Proposition 1 and Corollary 1 are clear proof sketches under a well-separated Gaussian-mixture assumption (Δ ≫ σ_max). A short empirical check of mode separation or mixture structure beyond the single Pursuit-Evasion observation in Fig. 3 would better connect the theory to the full dataset.
  2. [§IV-C.3; §I] DDIM with K=20 is motivated for real-time control, but no wall-clock latency, control frequency, or onboard timing numbers are reported. Even a brief timing table would support the deployment claim.
  3. [§III-B; Table I] Table I and the observation description are helpful; a one-sentence note on sensing radius relative to arena size (200×200) would make the information asymmetry more concrete for readers.
  4. [§II] Related work correctly distinguishes MIMIC-D (human-demonstration multi-agent diffusion) from privileged RL distillation; a slightly sharper sentence on what is new versus CTDS / interactive distillation would help novelty placement.
  5. [§IV-A; Fig. 1] Minor presentation: “discard” → “discarded” (§IV-A); arXiv IDs in related work for concurrent preprints are fine but should be updated if camera-ready versions exist; Fig. 1 caption is dense—consider splitting Stage 2/3 labels for readability.
  6. [§V-A; §VI] Scalability is only evaluated at N=3; the limitation is noted in §VI. A single N=5 or N=6 pilot (even if only Coverage) would reduce the risk that the multi-modality story is specific to small teams.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: mode-averaging corollary is standard MSE optimality under a stated mixture; empirical gains are external task metrics, not quantities defined by the training loss.

full rationale

This is an empirical methods paper whose central theoretical claim (Proposition 1 / Corollary 1) is a standard consequence of conditional-mean optimality of MSE under an assumed well-separated Gaussian mixture; it is not fitted to the experimental outcomes and does not redefine the reported metrics. Stage-2 supervision uses deterministic mean actions from successful collision-free oracle episodes only, but that is an acknowledged offline-distillation limitation (§VI), not a circular redefinition of success: the same filtered dataset is used for both CoDiMAD and BC-RNN, so Table II gaps isolate the effect of modeling the conditional distribution rather than its mean. Diffusion recovery is justified by the known score-matching property of DDPM (cited to Ho et al. and Song et al.), not by a self-citation uniqueness theorem. Task metrics (coverage rate, capture rate, box-pushing success, collisions) are external environment returns, not quantities constructed from the denoising loss. No prediction reduces by the paper’s equations to a fitted free parameter that is then re-sold as discovery. Self-citations (e.g., ARBoids) are background and not load-bearing for the multi-modality or distillation claims. Score 0 is therefore appropriate.

Assumptions & free parameters 4 free parameters · 5 assumptions · 1 invented entities

The central claim rests on standard Dec-POMDP and diffusion score-matching background, on MAPPO as a competent privileged teacher, and on several hand-chosen training hyperparameters. No new physical entities are postulated; the main modeling commitments are the multi-modal mixture idealization in Prop. 1 and the offline successful-trajectory filtering rule that defines the student data distribution.

free parameters (4)
  • Diffusion training steps T and DDIM inference steps K = T=200, K=20
    T=200 and K=20 are chosen by hand for quality vs latency; the claim that few-step DDIM remains multi-modal enough for coordination depends on this choice.
  • Cosine noise schedule and loss reweighting (1-ᾱ_k) = s=0.008, β∈[1e-4,0.02]
    Schedule parameters (s=0.008, β∈[1e-4,0.02]) and the (1-ᾱ_k) reweighting in Eq. (2) are design choices that shape which modes are learned.
  • Network and optimization hyperparameters = hidden=128, lr=1e-4, EMA=0.99, 50 epochs, batch=128
    GRU/MLP width 128, three residual blocks, AdamW lr=1e-4, EMA decay 0.99, 50 epochs, batch 128, 5M oracle env steps—standard but free knobs that affect reported gaps.
  • Oracle episode success/collision filter = keep only success & collision-free
    Only successful collision-free episodes enter D; this selection rule is a free methodological parameter that defines the supervised distribution.
assumptions (5)
  • domain assumption Multi-robot coordination is well-modeled as a Dec-POMDP with continuous 2D velocity actions and multi-channel BEV observations.
    §III-A–B; standard in multi-robot RL but shapes what “local observation multi-modality” means.
  • domain assumption A MAPPO oracle with dual-stream global maps learns near-optimal cooperative behavior that is worth cloning.
    Stage 1 and Table II oracle rows; if the oracle is poor, distillation cannot transfer good coordination.
  • standard math Ideal DDPM denoising recovers samples from the conditional data distribution p_D(a|o) via score matching (Ho et al., Song et al.).
    Remark 1 cites standard diffusion theory; practical finite-capacity/finite-step gap is acknowledged but not quantified.
  • ad hoc to paper Conditioned on a local observation, global states partition into well-separated configuration classes whose oracle actions form a Gaussian mixture with Δ ≫ σ_max.
    Proposition 1 assumptions; needed for the multi-modality and exponential low-density claims in Corollary 1.
  • ad hoc to paper Deterministic mean of the oracle Gaussian is an adequate action label for distillation (no need to store oracle samples).
    §IV-B: oracle rolled out by taking μ_πO; multi-modality then arises only from different globals mapping to the same local o.
invented entities (1)
  • CoDiMAD three-stage privileged diffusion distillation pipeline
    purpose: Package MAPPO oracle training, filtered offline local-action pairs, and conditional DDPM students as a deployable communication-free multi-robot method.
    Method-level construct rather than a physical entity; independent evidence is the reported sim experiments, not an external measurement channel.

how reviews work

0 comments
Cite this review

Pith. "Pith review of CoDiMAD: Diffusion-Based Privileged Distillation for Communication-Free Multi-Robot Coordination." pith.science (2026). https://pith.science/paper/73PV3ASR

@misc{pith2026260709587,
  author       = {Pith},
  title        = {Pith review of: CoDiMAD: Diffusion-Based Privileged Distillation for Communication-Free Multi-Robot Coordination},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/73PV3ASR}},
  note         = {Machine review of arXiv:2607.09587}
}
read the original abstract

Decentralized multi-robot coordination under partial observability remains challenging, especially in communication-free settings where agents must act solely from local sensor observations. Privileged policy distillation provides a promising approach by transferring knowledge from a globally informed oracle to sensor-constrained students. However, in multi-agent systems, the same local observation may correspond to multiple global configurations requiring qualitatively different cooperative actions, making the conditional action distribution inherently multi-modal. Standard deterministic distillation collapses these modes to their mean, often yielding invalid or hesitant actions. To address this issue, we propose CoDiMAD, a three-stage framework that trains a privileged oracle with MAPPO, constructs an offline dataset of local-observation-oracle-action pairs, and distills the oracle into decentralized students parameterized as conditional denoising diffusion probabilistic models. By approximating the conditional oracle-action distribution through the diffusion reverse process, CoDiMAD samples decisive actions from coherent coordination modes rather than averaging across them. Theoretical analysis characterizes the mode-averaging failure of deterministic distillation and the distributional recovery property of diffusion-based distillation. Experiments on three cooperative tasks show that CoDiMAD consistently outperforms direct local MARL and deterministic distillation baselines. The source code will be made publicly available upon acceptance.

Figures

Figures reproduced from arXiv: 2607.09587 by the authors.

Figure 1
Figure 1. Overview of the CoDiMAD framework. (i) Stage 1 (Oracle Policy Training): The oracle actor encodes dual-stream inputs—local maps o loc i and global maps o glo i —through parallel CNN branches, aggregates temporal features via a GRU, and produces actions at through an MLP head. A centralized MAPPO critic provides privileged value estimates during training. (ii) Stage 2 (Data Collection): The converged oracle policy is… view at source ↗
Figure 2
Figure 2. Overview of the three evaluation environments. (a) Coverage: agents cooperatively explore the arena to maximize grid coverage. (b) Pursuit￾Evasion: pursuers cooperate to capture a faster evader. (c) Box Pushing: agents push a heavy box to the goal region. rl denotes the LiDAR detection range, rc denotes the coverage radius, and re denotes the capture radius. Thus, in the well-separated regime m∆ ≫ σ, the density ass… view at source ↗
Figure 3
Figure 3. Conditional action distributions under partial observability. Under a fixed local observation oi in the Pursuit-Evasion task, the same local percept can be compatible with multiple global states and therefore multiple valid oracle actions. (a) Oracle actions from the 200 nearest-neighbor observations in the dataset, revealing two distinct coordination modes. (b) 200 independent CoDiMAD samples conditioned on the sam… view at source ↗
Figures from the paper (1 more)
Figure 4
Figure 4. Figure 4: Trajectory diversity from identical initial states. Five rollouts are shown for each method from the same initial condition. (a) and (b): Coverage. (c) and (d): Pursuit-Evasion. CoDiMAD produces diverse but individually coherent trajectories, corresponding to different…

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

31 extracted references · 3 linked inside Pith

  1. [1]

    Coped-advancing multi-robot collaborative perception: A comprehensive dataset in real- world environments,

    Y . Zhou, L. Quang, C. Nieto-Granda, and G. Loianno, “Coped-advancing multi-robot collaborative perception: A comprehensive dataset in real- world environments,”IEEE Robot. Automat. Lett., vol. 9, no. 7, 2024

    2024

  2. [2]

    H2gnn: Hierarchical-hops graph neural networks for multi-robot exploration in unknown environments,

    H. Zhang, J. Cheng, L. Zhang, Y . Li, and W. Zhang, “H2gnn: Hierarchical-hops graph neural networks for multi-robot exploration in unknown environments,”IEEE Robot. Automat. Lett., vol. 7, no. 2, pp. 3435–3442, 2022

    2022

  3. [3]

    Heterogeneous multi-robot cooperation with asynchronous multi-agent reinforcement learning,

    H. Zhang, X. Zhang, Z. Feng, and X. Xiao, “Heterogeneous multi-robot cooperation with asynchronous multi-agent reinforcement learning,” IEEE Robot. Automat. Lett., vol. 9, no. 1, pp. 159–166, 2024

    2024

  4. [4]

    F. A. Oliehoek and C. Amato,A concise introduction to decentralized POMDPs. Springer, 2016, vol. 1

    2016

  5. [5]

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

    R. Lowe, Y . Wu, A. Tamar, J. Harb, P. Abbeel, and I. Mordatch, “Multi- agent actor-critic for mixed cooperative-competitive environments,” in Proc. Adv. Neural Inf. Process. Syst., vol. 30, 2017, pp. 6379–6390

    2017

  6. [6]

    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,”J. Mach. Learn. Res., vol. 21, no. 178, pp. 1–51, 2020

    2020

  7. [7]

    ARBoids: Adaptive residual reinforcement learning with boids model for cooperative multi-USV target defense,

    J. Tao, T. Shen, D. Zhao, and F. Zhang, “ARBoids: Adaptive residual reinforcement learning with boids model for cooperative multi-USV target defense,”IEEE Robot. Automat. Lett., vol. 11, no. 3, pp. 3637– 3644, 2026

    2026

  8. [8]

    A survey of underwater multi-robot systems,

    Z. Zhou, J. Liu, and J. Yu, “A survey of underwater multi-robot systems,” IEEE/CAA J. Autom. Sin., vol. 9, no. 1, pp. 1–18, 2021

    2021

Show all 31 references
  1. [9]

    Multi-robot systems and cooperative object transport: Communications, platforms, and challenges,

    X. An, C. Wu, Y . Lin, M. Lin, T. Yoshinaga, and Y . Ji, “Multi-robot systems and cooperative object transport: Communications, platforms, and challenges,”IEEE Open J. Comput. Soc., vol. 4, pp. 23–36, 2023

    2023

  2. [10]

    A new learning paradigm: Learning using privileged information,

    V . Vapnik and A. Vashist, “A new learning paradigm: Learning using privileged information,”Neural Netw., vol. 22, no. 5-6, pp. 544–557, 2009

    2009

  3. [11]

    Learning by cheating,

    D. Chen, B. Zhou, V . Koltun, and P. Kr¨ahenb¨uhl, “Learning by cheating,” inProc. Conf. Robot Learn., 2020, pp. 66–75

    2020

  4. [12]

    RMA: Rapid motor adaptation for legged robots,

    A. Kumar, Z. Fu, D. Pathak, and J. Malik, “RMA: Rapid motor adaptation for legged robots,” inProc. Robot. Sci. Syst., 2021

    2021

  5. [13]

    Diffusion policy: Visuomotor policy learning via action diffusion,

    C. Chiet al., “Diffusion policy: Visuomotor policy learning via action diffusion,”Int. J. Robot. Res., vol. 44, no. 10-11, pp. 1684–1704, 2025

    2025

  6. [14]

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

    C. Yuet al., “The surprising effectiveness of PPO in cooperative multi- agent games,” inProc. Adv. Neural Inf. Process. Syst., vol. 35, 2022, pp. 24 611–24 624

    2022

  7. [15]

    Denoising diffusion probabilistic models,

    J. Ho, A. Jain, and P. Abbeel, “Denoising diffusion probabilistic models,” inProc. Adv. Neural Inf. Process. Syst., vol. 33, 2020, pp. 6840–6851

    2020

  8. [16]

    Denoising diffusion implicit models,

    J. Song, C. Meng, and S. Ermon, “Denoising diffusion implicit models,” inProc. Int. Conf. Learn. Represent., 2021

    2021

  9. [17]

    Learning multiagent commu- nication with backpropagation,

    S. Sukhbaatar, A. Szlam, and R. Fergus, “Learning multiagent commu- nication with backpropagation,” inProc. Adv. Neural Inf. Process. Syst., vol. 29, 2016, pp. 2244–2252

    2016

  10. [18]

    Learning to communicate with deep multi-agent reinforcement learning,

    J. N. Foerster, Y . M. Assael, N. de Freitas, and S. Whiteson, “Learning to communicate with deep multi-agent reinforcement learning,” inProc. Adv. Neural Inf. Process. Syst., vol. 29, 2016

    2016

  11. [19]

    TarMAC: Targeted multi-agent communication,

    A. Daset al., “TarMAC: Targeted multi-agent communication,” inProc. Int. Conf. Mach. Learn., 2019, pp. 1538–1546

    2019

  12. [20]

    Learning attentional communication for multi-agent cooperation,

    J. Jiang and Z. Lu, “Learning attentional communication for multi-agent cooperation,” inProc. Adv. Neural Inf. Process. Syst., vol. 31, 2018, pp. 7254–7264

    2018

  13. [21]

    Off- belief learning,

    H. Hu, A. Lerer, B. Cui, L. Pineda, N. Brown, and J. Foerster, “Off- belief learning,” inProc. Int. Conf. Mach. Learn., 2021, pp. 4369–4379

    2021

  14. [22]

    Heterogeneous-agent reinforcement learning,

    Y . Zhong, J. G. Kuba, X. Feng, S. Hu, J. Ji, and Y . Yang, “Heterogeneous-agent reinforcement learning,”J. Mach. Learn. Res., vol. 25, no. 32, pp. 1–67, 2024

    2024

  15. [23]

    CTDS: Cen- tralized teacher with decentralized student for multiagent reinforcement learning,

    J. Zhao, X. Hu, M. Yang, W. Zhou, J. Zhu, and H. Li, “CTDS: Cen- tralized teacher with decentralized student for multiagent reinforcement learning,”IEEE Trans. Games, vol. 16, no. 1, pp. 140–150, 2022

    2022

  16. [24]

    KD-MARL: Resource-aware knowledge distillation in multi-agent reinforcement learning,

    C. Tanget al., “KD-MARL: Resource-aware knowledge distillation in multi-agent reinforcement learning,”arXiv preprint arXiv:2604.06691, 2025

    2025 arXiv

  17. [25]

    Interactive distillation for cooperative multi-agent reinforcement learning,

    M. Cho, B. Altundas, and M. Gombolay, “Interactive distillation for cooperative multi-agent reinforcement learning,”arXiv preprint arXiv:2601.05407, 2026

    2026

  18. [26]

    Planning with diffusion for flexible behavior synthesis,

    M. Janner, Y . Du, J. B. Tenenbaum, and S. Levine, “Planning with diffusion for flexible behavior synthesis,” inProc. Int. Conf. Mach. Learn., 2022, pp. 9902–9915

    2022

  19. [27]

    MIMIC-D: Multi-modal imitation for multi-agent coordination with decentralized diffusion policies,

    N. Negarmehr, N. Mehret al., “MIMIC-D: Multi-modal imitation for multi-agent coordination with decentralized diffusion policies,”arXiv preprint arXiv:2509.14159, 2025

    2025 arXiv

  20. [28]

    Learning phrase representations using rnn encoder– decoder for statistical machine translation,

    K. Choet al., “Learning phrase representations using rnn encoder– decoder for statistical machine translation,” inProc. Conf. Empirical Methods Nat. Lang. Process., 2014, pp. 1724–1734

    2014

  21. [29]

    A connection between score matching and denoising au- toencoders,

    P. Vincent, “A connection between score matching and denoising au- toencoders,”Neural Comput., vol. 23, no. 7, pp. 1661–1674, 2011

    2011

  22. [30]

    Score-based generative modeling through stochastic differ- ential equations,

    Y . Song, J. Sohl-Dickstein, D. P. Kingma, A. Kumar, S. Ermon, and B. Poole, “Score-based generative modeling through stochastic differ- ential equations,” inProc. Int. Conf. Learn. Represent., 2021

    2021

  23. [31]

    IDQL: Implicit q-learning as an actor-critic method with diffusion policies,

    P. Hansen-Estruch, I. Kostrikov, M. Janner, J. G. Kuba, and S. Levine, “IDQL: Implicit q-learning as an actor-critic method with diffusion policies,”arXiv preprint arXiv:2304.10573, 2023

    2023 arXiv

Pith tools

Reviewed July 13, 2026 · model on record in the stance chip above.