Pith. sign in

REVIEW 5 major objections 6 minor 27 references

Curriculum Generation under Structured Parametric Environments for Robust Navigation Policies

T0 review · 5 major / 6 minor · reviewed 2026-08-14 · deepseek-v4-flash

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

Pith's one-line read This paper claims that a value-network critic can tune environment difficulty bounds by gradient ascent, producing navigation policies that generalize better than those trained with hand-made or stochastic curricula.

desk verdict A clearly derived reparameterized curriculum critic whose headline empirical claim is not supported by its own per-seed tables; the core idea is worth engaging but the paper needs major revision. read the letter →

arxiv 2608.08545 v1 pith:ECJ2R3OL submitted 2026-08-09 cs.RO cs.LG

Prishita Ray This is my paper · ORCID
classification cs.ROcs.LG
keywords CurriculumGenerationReinforcementLearningRobustNavigationContinuousControlReparameterizationAutomaticPPO
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 claims that in continuous-parameter environments, curriculum generation can be done by direct gradient updates on environment difficulty bounds rather than by sampling tasks from hand-designed or competence-based distributions. The proposal is a reparameterized curriculum in which the policy is trained with Proximal Policy Optimization while a value network acts as curriculum critic: environments are sampled uniformly inside a running range, the sampling is reparameterized so the upper bound is differentiable, and the bound is raised only when raising it would lower the policy's estimated value. The authors report that this mechanism, sometimes with an added distribution-shift penalty, consistently beats vanilla training, uniform random sampling, manual linear curricula, and the SPRL and ALP-GMM automatic-curriculum baselines across five seeds in modified Car Racing and Bipedal Walker environments. If true, the result means a simple, cheap gradient rule can replace stochastic task selection in structured environment spaces, and the main cost is knowing the environment parameters at deployment rather than engineering the curriculum.

What carries the argument

The load-bearing object is the reparameterized curriculum bound: $\psi_i = \psi_{\min} + z_i \odot (\psi_t - \psi_{\min})$ with $z_i \sim U[0,1]$, making the sampled environment a linear function of the running upper bound. This turns the curriculum update into a one-line gradient rule: $\frac{\partial L}{\partial \psi_t[j]} = \frac{1}{n}\sum_i \left(\frac{\partial V_\phi(x_i, \psi_i)}{\partial \psi_i[j]} - 2\beta(\psi_i[j] - \psi_{\text{prev},i}[j])\right) z_i[j]$, with the bound increased only when the gradient is negative, meaning only when difficulty growth would lower estimated value. The value network doubles as the curriculum critic, so no separate teacher or black-box optimizer is needed; the distribution-shift term $\delta = \|\psi_i - \psi_{\text{prev},i}\|_2$ with weight $\beta$ is an optional regularizer that the paper finds useful in the image-based Car Racing setting.

What would settle it

Train the same method with the parameter conditioning inputs $\psi$ replaced by a constant or by a deliberately wrong estimate while keeping the curriculum update; if the final test-reward distribution over the full parameter range is unchanged, then the conditioning mechanism is not what does the work. Alternatively, run the curriculum update with the gradient sign flipped (shrink bounds when value is low) and identical compute; if the flipped version matches the reported performance, the gradient direction is not causal.

Watch

Extended reading notes

Core claim

The paper's central claim is that the difficulty of a training environment can be treated as a differentiable parameter of the value function, and that moving difficulty upward along the value gradient produces a robust policy. Concretely, with environment parameters $\psi$ bounded by a lower vector $\psi_{\min}$ and a running upper vector $\psi_t$, the authors sample environments as $\psi_i = \psi_{\min} + z_i \odot (\psi_t - \psi_{\min})$, giving $\partial \psi_i / \partial \psi_t = z_i$; the value network $V_\phi$ then yields a critic objective whose negative gradient indicates that increasing an upper bound would reduce expected policy value. Bounds are updated only when that gradient is negative — "unidirectional" curriculum growth — with learning rates annealed and steps normalized by the parameter range. The paper reports that this scheme, without adversarial or population-based search, outperforms six baselines on the two benchmark tasks, with the MSE-regularized variant best on the image-based Car Racing task and the unregularized variant best on the proprioceptive Bipedal Walker task.

Load-bearing premise

The policy is given the ground-truth environment parameters $\psi$ as part of its observation, and the curriculum update assumes that higher parameter values always mean harder environments; if those parameters are unavailable or estimated incorrectly at test time, the reported robustness will not transfer.

Editorial extensions

If this is right

  • At equal training time, the gradient-based curriculum reaches higher mean episodic reward than stochastic baselines on the two tasks.
  • Unidirectional updates (only raise difficulty) outperform bidirectional updates that can lower bounds, because bidirectional updates slow learning.
  • The distribution-shift regularization (MSE) improves the multimodal CNN+MLP policy in Car Racing but does not help the MLP-only Bipedal Walker policy.
  • Reverse curriculum and frontier sampling underperform the forward running-range curriculum, suggesting that sampling the full growing range with a gradient critic is a better exploration schedule.
  • Training time stays comparable to vanilla and baseline methods, so the improvement is not bought with extra compute.

Reading between the lines

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

  • One implication not drawn by the authors: the reparameterization trick used here is the same interface as in variational inference, so any differentiable environment simulator with bounded continuous knobs could be tuned by this rule without a separate teacher network.
  • A testable extension would replace the ground-truth $\psi$ conditioning with a learned encoder that estimates $\psi$ from observations; if the estimator is accurate enough, the robustness should survive and the method would work in partially observed settings.
  • The paper assumes higher parameter values monotonically increase difficulty, but never verifies it; a curriculum that also learns the direction of difficulty from the value gradient might be a more general recipe for real environments where difficulty is non-monotonic.
  • Compared with adversarial minimax environment design, this single-step gradient ascent on the value function may be a cheaper alternative when the parameter space is low-dimensional and structured, though the paper does not test that comparison directly.
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

5 major / 6 minor

Summary. The paper proposes Reparam, a curriculum-generation method for reinforcement learning in environments with continuous parameter vectors. A PPO value network is used as a critic; environment parameters are sampled via a reparameterized uniform distribution over [ψ_min, ψ_t], and gradients of the critic objective with respect to the upper-bound vector ψ_t are used to move the curriculum. An optional distribution-shift regularizer is introduced, and updates are restricted to directions that increase difficulty ('unidirectional'). The method is evaluated on modified Car Racing (κ, p) and Bipedal Walker (f, ρ, s) benchmarks against vanilla, random, manual, SPRL, ALP-GMM, frontier, and reverse-curriculum baselines over five seeds, with per-seed and IQM tables in the appendices. The central claim is that the method consistently outperforms all baselines.

Significance. If the comparative claims held, the paper would offer a simple, interpretable alternative to black-box curriculum search: a reparameterized gradient update over continuous parameter bounds plus an optional regularizer. The Appendix B derivation is algebraically correct, and the authors are transparent in reporting per-seed results and curriculum growth curves. The main weakness is that the evidence as presented does not support the headline 'consistently outperforms' claim, and the sign convention and monotonicity assumptions underlying the update need clarification. The contribution is potentially interesting but requires substantial additional analysis before it can be accepted.

major comments (5)
  1. [Section 5.4, Appendix Tables 5 and 6] The abstract claims that the method 'consistently outperforms' vanilla, random, manual, SPRL, ALP-GMM, frontier, and reverse-curriculum baselines across five seeds, but the per-seed tables do not support this. In Bipedal Walker (Table 6), Reparam beats Manual on only seeds 3 and 4 (136 vs 37 and 152 vs 121) and loses on seeds 0, 1, and 2; against ALP-GMM it wins 3 of 5 seeds, and all per-seed standard deviations overlap. In Car Racing (Table 5), the unregularized Reparam variant obtains 448±122, below Vanilla (540±178) and Manual (617±166), so only Reparam-M outperforms, and even there Manual wins seed 1 (702 vs 699). No significance tests, confidence intervals, or multiple-comparison corrections are reported, and the reported means are obtained by selecting the best checkpoint per seed. The headline claim should be softened or supported by paired significance testing across seeds.
  2. [Section 4.2, Eq. (6)] The update rule increases ψ_t only when ∂L/∂ψ_t < 0, that is, when the critic predicts that raising the upper bound reduces estimated value. This is the opposite of a competence-based self-paced curriculum, in which difficulty should grow when the policy is competent; as written, it is an adversarial value-minimizing update. If the intent is to make the curriculum adversarial, the method should be framed and compared against UED/minimax baselines; if the intent is self-paced, the condition appears backwards. The paper should clarify the objective and justify the sign convention, and ideally validate the gradient against a Monte-Carlo estimate of returns, since the value network is itself a moving target during training.
  3. [Section 3.2, Section 5.4, Eq. (6)] The curriculum grows all upper bounds, including friction f in Bipedal Walker. The paper explicitly states the monotonic difficulty assumption only for Car Racing (§3.1: higher κ and p correspond to more difficult tracks). For f, higher friction generally makes locomotion easier (less slipping), so the sign of difficulty may be inverted. No calibration is provided (for example, reward of a fixed policy across a grid of (f, ρ, s) values). Since the update direction in Eq. (6) depends on the assumption that increasing any bound increases difficulty, this needs to be verified; otherwise the method may be moving in the wrong direction.
  4. [Section 5.2, Tables 3 and 4] The advanced variants Reparam-R and Frontier are configured with β=1.0 in Car Racing but β=0.0 in Bipedal Walker, 'based on the best-performing Reparam configuration' (§5.2). This is a post-hoc per-environment selection of baseline hyperparameters using knowledge of the evaluation outcomes, which biases the comparison in favor of the proposed variants. Please fix the baseline configurations before evaluation or report a sensitivity analysis over β for every variant, rather than choosing the best-performing configuration after the fact.
  5. [Sections 3.1, 3.2, 4.1, Figure 2] The policy is conditioned on the ground-truth environment parameters ψ=(κ, p) or (f, ρ, s) as part of the observation. If these parameters are not available at deployment, the trained policy cannot use its conditioning mechanism, so the reported robustness gains do not transfer. The paper motivates the work with robust navigation policies for autonomous agents, but does not discuss how ψ would be obtained in practice or evaluate with estimated, noisy, or partial parameters. This limitation should be stated explicitly, and the claims should be scoped to settings where the ground-truth parameters are known.
minor comments (6)
  1. [Appendix D and E] The appendix figure numbers duplicate the main-text figures (Appendix D has Figures 2 and 3, Appendix E has Figures 4 and 5); the appendices should be renumbered.
  2. [Section 4.2] The notation δ(ψ_i, ψ_prev_i) is not fully defined: ψ_prev_i is not specified as a vector equal to the previous upper bound or as a sampled environment, and the regularization is described as a distribution shift while the formula is a per-sample Euclidean distance.
  3. [Equation (1)] The term 'N denotes the number of delays' is used without defining what constitutes a delay in the Car Racing environment.
  4. [Tables 1 and 2] The captions report mean±standard deviation, but it is not clear whether the standard deviation is over the 500 test environments or across the five training seeds; the per-seed tables in the appendix suggest the former, so the caption should state this explicitly.
  5. [Appendix D and E] The abbreviation IQM is used without definition; the interquartile mean should be defined at first use in the main text.
  6. [Section 5.3] The paper states that hyperparameters are tuned manually; a sensitivity analysis for the annealing rate α, the regularization weight β, and the initial bounds would strengthen the robustness claims.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the curriculum update is a heuristic guided by the PPO value network, and the headline comparison is evaluated on independently sampled test environments, not on the quantities used to fit the curriculum.

full rationale

The paper's reparameterized curriculum method computes gradients of a critic objective L=1/n sum (V_phi(x_i, psi_i) - beta*delta) with respect to curriculum upper bounds psi_t, then increases a bound only when the gradient is negative. This is a closed-loop algorithmic feedback loop: the critic is the PPO value network, which is trained on the same policy and environment distribution it is guiding. However, this is not a logical circularity. The paper does not claim the critic gradient is an external 'prediction' or 'first-principles result'; it is an explicit design choice for adapting the training distribution. The central empirical claim, that Reparam/Reparam-M outperforms baselines, is evaluated by selecting a checkpoint using held-out evaluation environments and re-evaluating on 500 independently sampled test environments from the full parameter range (Section 5.3). The final comparison is therefore not forced by construction: a poor critic would produce a poor curriculum and the method could lose to baselines, which is exactly what happens in several per-seed results. The only self-citation is reference [3] (Banerjee, Ray, and Campbell) in the related-work discussion of Bayesian optimization; it is not load-bearing for any of the paper's claims. The statistical weakness of the 'consistently outperforms' claim, and the deployment assumption that ground-truth environment parameters are available, are correctness or applicability concerns, not circularity. Under the hard rule requiring a specific reduction to inputs, no circular step can be identified.

Assumptions & free parameters 4 free parameters · 5 assumptions · 0 invented entities

The method's load-bearing assumptions are: ground-truth environment parameters are supplied to the policy, higher parameter values are monotonically harder, the in-training value network gives trustworthy gradient directions for the curriculum, and a unidirectional increase-only rule is sufficient. The annealing rate alpha is unreported, and beta and initial bounds are manually chosen.

free parameters (4)
  • annealing rate alpha = not reported
    Equation 3 defines the annealed learning rate as lr / alpha^t, but alpha is absent from the hyperparameter tables; the curriculum schedule depends on it.
  • beta regularization weight = 0.0 for Reparam and Reparam-A in Bipedal Walker, 1.0 for Reparam-M in both environments and Reparam-A in Car Racing
    Beta is manually tuned and gates the distribution-shift regularization term delta in Eq. 5; it is a central component of the regularized variant that reports the best Car Racing result.
  • initial curriculum upper bounds psi_t0 = Car Racing: kappa=0.32, p=0.051; Bipedal Walker: f=2.1, rho=0.02, s=0.001
    Hand-chosen starting points (Appendix C, Tables 3 and 4) determine how fast and how far the curriculum grows.
  • unidirectional update condition threshold = gradient < 0
    The choice to update psi_t only when the value gradient is negative is a design decision that defines the curriculum; no convergence or optimality justification is provided.
assumptions (5)
  • domain assumption The environment parameter vector psi is observable and is concatenated into the policy input.
    Sections 3.1, 3.2, and 4.1 state the policy is conditioned on psi; this is not discussed as a deployment limitation.
  • domain assumption Higher values of kappa, p, f, rho, and s correspond monotonically to more difficult environments.
    Section 3.1 says 'higher values correspond to more difficult tracks'; the curriculum direction relies on this monotonicity, which is not verified.
  • domain assumption The PPO value network V_phi provides a reliable estimate of policy value for guiding curriculum bound updates.
    Section 4.2 differentiates V_phi through reparameterized samples; value estimates from an in-training critic are noisy and distribution-dependent.
  • ad hoc to paper Only increasing the upper curriculum bounds is sufficient to produce robust policies.
    The unidirectional update rule (Eq. 6) is an explicit design choice; the paper gives no convergence argument, and ablation Reparam-A (bidirectional) is worse in their runs.
  • standard math The reparameterized uniform sampling derivative partial(psi_i)/partial(psi_t) = z_i is valid.
    Appendix B derives the chain rule step correctly; this is standard reparameterization and does not introduce new assumptions.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Curriculum Generation under Structured Parametric Environments for Robust Navigation Policies." pith.science (2026). https://pith.science/paper/ECJ2R3OL

@misc{pith2026260808545,
  author       = {Pith},
  title        = {Pith review of: Curriculum Generation under Structured Parametric Environments for Robust Navigation Policies},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/ECJ2R3OL}},
  note         = {Machine review of arXiv:2608.08545}
}
read the original abstract

Robust navigation policies for autonomous agents must generalize across continuously varying environmental conditions such as turn rates, obstacles, friction, pits, and slopes. Curriculum generation provides a principled mechanism for improving generalization by progressively adapting training environments, but designing such curricula in a sample-efficient and automated manner remains challenging. This paper proposes a reparameterized curriculum generation framework for structured continuous environment parameters using unidirectional gradient-based optimization. To improve robustness in multimodal observation spaces consisting of image-based and scalar inputs, a distribution-shift regularization objective is incorporated to encourage the learning of finer-grained latent representations. The proposed method is evaluated across two continuous-control OpenAI Gym environments: a 2D obstacle-based Car Racing variant and Bipedal Walker variant, where coupled environment parameters jointly influence policy performance. Across five random seeds, our method consistently outperforms vanilla policy training, random parameter sampling, manual curricula, frontier-based methods, Self-Paced Reinforcement Learning (SPRL), Absolute Learning Progress with Gaussian Mixture Models (ALP-GMM), and reverse curriculum learning baselines. Ablation studies further demonstrate the effectiveness of the reparameterized curriculum mechanism across both environments, while highlighting environment-dependent benefits of the auxiliary regularization objective.

Figures

Figures reproduced from arXiv: 2608.08545 by the authors.

Figure 1
Figure 1. Examples of Car Racing and Bipedal Walker environments under varying parameter set [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Overview of the Reparam framework. The policy network (green) is optimized using PPO, [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. Mean Evaluation performance measured every 10000 training timesteps; shaded regions [PITH_FULL_IMAGE:figures/full_fig_p007_3.png] view at source ↗
Figures from the paper (6 more)
Figure 4
Figure 4. Figure 4: Curriculum Growth Curves in a) Car Racing and b) Bipedal Walker [PITH_FULL_IMAGE:figures/full_fig_p007_4.png]
Figure 5
Figure 5. Figure 5: Architecture of the Policy and Critic Networks in Car Racing and Bipedal Walker Envi [PITH_FULL_IMAGE:figures/full_fig_p011_5.png]
Figure 2
Figure 2. Figure 2: Mean evaluation performance measured every 10000 training timesteps. Curves show the [PITH_FULL_IMAGE:figures/full_fig_p014_2.png]
Figure 3
Figure 3. Figure 3: Curriculum Growth Curves and Sampling Ranges- Car Racing [PITH_FULL_IMAGE:figures/full_fig_p017_3.png]
Figure 4
Figure 4. Figure 4: Mean evaluation performance measured every 10000 training timesteps. Curves show the [PITH_FULL_IMAGE:figures/full_fig_p019_4.png]
Figure 5
Figure 5. Figure 5: Curriculum Growth Curves and Sampling Ranges- Bipedal Walker [PITH_FULL_IMAGE:figures/full_fig_p022_5.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

27 extracted references · 23 canonical work pages

  1. [1]

    Bengio, J

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

    2009

  2. [2]

    Sullivan, R

    R. Sullivan, R. P ´egoud, A. U. Rahmen, X. Yang, J. Huang, A. Verma, N. Mitra, and J. P. Dickerson. Syllabus: Portable curricula for reinforcement learning agents.Reinforcement Learning Journal, 2025

    work page 2025

  3. [3]

    Improving Environment Robustness of Deep Reinforcement Learning Approaches for Autonomous Racing Using Bayesian Optimization-based Curriculum Learning

    R. Banerjee, P. Ray, and M. Campbell. Improving environment robustness of deep reinforce- ment learning approaches for autonomous racing using bayesian optimization-based curricu- lum learning.arXiv preprint arXiv:2312.10557, 2023

    work page Pith review arXiv 2023

  4. [4]

    Jiang, M

    M. Jiang, M. Dennis, J. Parker-Holder, J. Foerster, E. Grefenstette, and T. Rockt¨aschel. Replay- guided adversarial environment design.Advances in Neural Information Processing Systems, 34:1884–1897, 2021

    work page 2021

  5. [5]

    A. S. Azad, I. Gur, J. Emhoff, N. Alexis, A. Faust, P. Abbeel, and I. Stoica. Clutr: Curricu- lum learning via unsupervised task representation learning. InInternational Conference on Machine Learning, pages 1361–1395. PMLR, 2023

    work page 2023

  6. [6]

    C. Wang, J. Wang, Y . Shen, and X. Zhang. Autonomous navigation of uavs in large-scale com- plex environments: A deep reinforcement learning approach.IEEE Transactions on V ehicular Technology, 68(3):2124–2136, 2019

    work page 2019

  7. [7]

    Song and J

    Y . Song and J. Schneider. Robust reinforcement learning via genetic curriculum. In2022 International Conference on Robotics and Automation (ICRA), pages 5560–5566, 2022

    work page 2022

  8. [8]

    Narvekar, B

    S. Narvekar, B. Peng, M. Leonetti, J. Sinapov, M. E. Taylor, and P. Stone. Curriculum learning for reinforcement learning domains: A framework and survey.Journal of Machine Learning Research, 21(181):1–50, 2020

    work page 2020

Show all 27 references
  1. [9]

    Portelas, C

    R. Portelas, C. Colas, L. Weng, K. Hofmann, and P.-Y . Oudeyer. Automatic curriculum learn- ing for deep rl: A short survey.arXiv preprint arXiv:2003.04664, 2020

    2003 arXiv

  2. [10]

    Matiisen, A

    T. Matiisen, A. Oliver, T. Cohen, and J. Schulman. Teacher-student curriculum learning. In IEEE Transactions on Neural Networks and Learning Systems, 2019

    2019

  3. [11]

    R. Wang, J. Lehman, J. Clune, and K. O. Stanley. Paired open-ended trailblazer (poet): End- lessly generating increasingly complex and diverse learning environments and their solutions. arXiv preprint arXiv:1901.01753, 2019

    1901 arXiv

  4. [12]

    Parker-Holder, M

    J. Parker-Holder, M. Jiang, M. Dennis, M. Samvelyan, J. Foerster, E. Grefenstette, and T. Rockt¨aschel. Evolving curricula with regret-based environment design. InInternational Conference on Machine Learning, pages 17473–17498. PMLR, 2022

    2022

  5. [13]

    Dennis, N

    M. Dennis, N. Jaques, E. Vinitsky, A. Bayen, S. Russell, A. Critch, and S. Levine. Emergent complexity and zero-shot transfer via unsupervised environment design.Advances in neural information processing systems, 33:13049–13061, 2020

    2020

  6. [14]

    A. Z. Ren and A. Majumdar. Distributionally robust policy learning via adversarial environ- ment generation.IEEE Robotics and Automation Letters, 7(2):1379–1386, 2022

    2022

  7. [15]

    Cobbe, C

    K. Cobbe, C. Hesse, J. Hilton, and J. Schulman. Leveraging procedural generation to bench- mark reinforcement learning. InInternational Conference on Machine Learning (ICML), 2020

    2020

  8. [16]

    Anzalone, S

    L. Anzalone, S. Barra, and M. Nappi. Reinforced curriculum learning for autonomous driving in carla. In2021 IEEE International Conference on Image Processing (ICIP), pages 3318–

  9. [17]

    Z. Qiao, K. Muelling, J. M. Dolan, P. Palanisamy, and P. Mudalige. Automatically generated curriculum based reinforcement learning for autonomous vehicles in urban environment. In 2018 IEEE Intelligent V ehicles Symposium (IV), pages 1233–1238. IEEE, 2018

    2018

  10. [18]

    K. Ryu, Q. Liao, Z. Li, P. Delgosha, K. Sreenath, and N. Mehr. Curricullm: Automatic task curricula design for learning complex robot skills using large language models. In2025 IEEE International Conference on Robotics and Automation (ICRA), pages 4470–4477. IEEE, 2025

    2025

  11. [19]

    Pinto, J

    L. Pinto, J. Davidson, R. Sukthankar, and A. Gupta. Robust adversarial reinforcement learning. InInternational Conference on Machine Learning (ICML), 2017

    2017

  12. [20]

    Vinitsky, Y

    E. Vinitsky, Y . Du, K. Parvate, K. Jang, P. Abbeel, and A. Bayen. Robust reinforcement learning using adversarial populations.arXiv preprint arXiv:2008.01825, 2020

    2008 arXiv

  13. [21]

    J. Moos, K. Hansel, H. Abdulsamad, S. Stark, D. Clever, and J. Peters. Robust reinforcement learning: A review of foundations and recent advances.Machine Learning and Knowledge Extraction, 4(1):276–315, 2022

    2022

  14. [22]

    Klink, C

    P. Klink, C. D’Eramo, J. R. Peters, and J. Pajarinen. Self-paced deep reinforcement learning. Advances in Neural Information Processing Systems, 33:9216–9227, 2020

    2020

  15. [23]

    Klink, H

    P. Klink, H. Yang, C. D’Eramo, J. Peters, and J. Pajarinen. Curriculum reinforcement learning via constrained optimal transport. InInternational Conference on Machine Learning, pages 11341–11358. PMLR, 2022

    2022

  16. [24]

    Portelas, C

    R. Portelas, C. Colas, K. Hofmann, and P.-Y . Oudeyer. Teacher algorithms for curriculum learning of deep rl in continuously parameterized environments. InConference on Robot Learning, pages 835–853. PMLR, 2020

    2020

  17. [25]

    Jiang, E

    M. Jiang, E. Grefenstette, and T. Rockt ¨aschel. Prioritized level replay. InInternational Con- ference on Machine Learning, pages 4940–4950. PMLR, 2021

    2021

  18. [26]

    Shukla, C

    Y . Shukla, C. Thierauf, R. Hosseini, G. Tatiya, and J. Sinapov. Acute: Automatic curriculum transfer from simple to complex environments.arXiv preprint arXiv:2204.04823, 2022

    2022 arXiv

  19. [27]

    Tobin, R

    J. Tobin, R. Fong, A. Ray, J. Schneider, W. Zaremba, and P. Abbeel. Domain randomization for transferring deep neural networks from simulation to the real world. InIEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2017. 10 Appendix A Policy-Critic Arc...

    2017

Pith tools

Reviewed August 14, 2026 · model on record in the stance chip above.