Pith. sign in

REVIEW 5 major objections 6 minor 21 references

Reinforcement learning with world model

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

Pith's one-line read The paper asserts that an agent combining Soft Actor-Critic with a recurrent world model and a decaying curiosity bonus beats prior model-free methods on MuJoCo control and loses less performance under partial observation.

desk verdict A plausible SAC+RNN combination that is undone by an unbuildable curiosity reward and unsupported SOTA claims. read the letter →

arxiv 1908.11494 v4 pith:QJP54ODN submitted 2019-08-30 cs.AI

classification cs.AI
keywords reinforcementlearningworldmodelPOMDPcuriosityrecurrentneuralnetworkSoftActor-Criticsampleefficiencycontinuouscontrol
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 argues that an intelligent reinforcement-learning agent needs intuition, memory, prediction, and curiosity, and that these can be combined in one architecture. It proposes RMC, which pairs Soft Actor-Critic as the decision-making intuition with a recurrent network that learns to predict the next state and supplies a curiosity reward from prediction error. The authors claim RMC reaches new state-of-the-art results on MuJoCo continuous-control benchmarks while keeping sample efficiency and stable training, and that the same agent transfers to partially observed POMDP tasks without reconstructing hidden states by hand. If true, this would mean model-free agents can gain the benefits of a learned world model without sacrificing off-policy sample efficiency, and can tolerate missing observations more gracefully.

What carries the argument

The load-bearing object is the recurrent hidden state $s_t$ produced by a GRU from the previous hidden state, the current observation, and the previous action. The paper trains this hidden state with two objective terms: a model head minimizing the residual L1 prediction error $\|s_{t+1} - \hat f_\psi(s_t, a_t)\|$ and a value head providing the soft Q Bellman loss, while gradients from the policy head are blocked. The same prediction error supplies the intrinsic curiosity reward, scaled by a decaying $\beta$ and added to the extrinsic reward. This design is meant to push the learned $s_t$ toward a sufficient statistic of the environment so that Soft Actor-Critic's policy can act as if the state were fully observed.

What would settle it

Train RMC and SAC with the same network sizes, replay settings, and seeds on all five MuJoCo environments, recording final return and total environment steps; the central claim is falsified if RMC does not exceed SAC in final return or reach the same return in fewer steps on most tasks. For the POMDP claim, evaluate an MDP-trained RMC at flicker probabilities from 0.1 to 0.9 and compare the normalized score drop with MDP-trained SAC; the paper predicts a strictly smaller drop at every probability.

Watch

Extended reading notes

Core claim

On its own terms, the paper's discovery is that one recurrent encoder can serve three roles at once: memory through a GRU hidden state, prediction through an L1 residual next-state model, and curiosity through a decaying prediction-error bonus, while the SAC policy acts purely on the learned hidden state. The authors report that RMC outperforms SAC, TD3, PPO, and DDPG on MuJoCo tasks, outperforms SAC with frame stacking on the flickering MuJoCo POMDP, and preserves more performance than SAC when agents trained on MDPs are evaluated under increasing levels of observation obscuring. They attribute the result to the choice of which losses shape the recurrent representation: model loss plus value loss are back-propagated into the RNN, while policy gradients are deliberately blocked for training stability.

Load-bearing premise

The plan works only if the hidden state formed by jointly predicting the next state and estimating future reward is a good enough substitute for the true environment state in every task, yet the paper demonstrates this representation choice on one ablation environment, HalfCheetah-v2.

Editorial extensions

If this is right

  • If RMC's central claim holds, an off-policy agent can incorporate a recurrent world model without sacrificing sample efficiency, outperforming SAC, TD3, PPO, and DDPG on MuJoCo control tasks.
  • In the flickering MuJoCo POMDP, a single observation per time step plus a recurrent hidden state beats SAC with frame stacking, so memory can substitute for explicit observation history.
  • The best of the six training schemes is to back-propagate model loss and value loss into the RNN while blocking policy gradients, which prescribes how to combine supervised world-model learning with reinforcement learning.
  • A decaying curiosity bonus based on model prediction error improves early exploration, while setting it to zero or too large hurts performance, supporting the use of a scheduled intrinsic reward.
  • Because the MDP-trained RMC degrades less than SAC when observations are obscured, the framework suggests partially observed control tasks can be tackled without hand-designed hidden-state inference.

Reading between the lines

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

  • A testable extension the paper does not pursue is to replace SAC with another off-policy learner such as TD3 or DDPG; if the recurrent representation is the real driver, RMC's gains should persist, separating the contribution of the world model from the contribution of SAC.
  • The paper reports no performance loss at flicker probability $p=0.5$; evaluating at higher obscuration probabilities or with longer occlusions would show whether the recurrent state truly stores missing information or only tolerates occasional blank frames.
  • The curiosity bonus is tested only as a decaying scalar on prediction error; an adaptive, state-dependent bonus could keep exploration alive in sparse-reward environments without the need for a fixed decay schedule.
  • If the learned hidden state is a sufficient statistic, the same encoder could be reused for transfer between MDP and POMDP versions of a task, which is a direct prediction the paper leaves untested.
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 RMC (Recurrent Model with Curiosity), an actor-critic agent that combines Soft Actor-Critic with an RNN-based world model. The world model is trained jointly with the critic to produce a latent state representation, and the prediction error is used to define an intrinsic curiosity reward intended to encourage exploration. The authors claim state-of-the-art results on several MuJoCo continuous-control tasks, improved robustness on a flickering MuJoCo POMDP, and analyze ablations over six backpropagation schemes and over the curiosity reward weight. The main contribution is the architecture and the joint-training scheme, but the presentation is incomplete and the algorithm as written contains several internal inconsistencies that undermine the reported experiments.

Significance. If the experimental claims were substantiated, RMC would be a useful demonstration that a recurrent world model with a joint model/value representation can improve sample efficiency and partially observable robustness over strong model-free baselines. The architectural idea of blocking policy gradients into the RNN and using both model and value losses for representation learning is interesting and worth investigating. However, the paper does not provide quantitative results, error bars, or statistical tests, and the algorithm description has multiple errors: the curiosity reward is never added to any stored reward or training objective, the policy loss is inverted relative to SAC, and the target-network update in Algorithm 1 is a no-op. These are not presentation issues; they mean the method as written cannot reproduce the claimed results. Without code or complete hyperparameter details, the central claims are not verifiable.

major comments (5)
  1. [§3.3, §3.4, Algorithm 1] The curiosity reward is not implemented in any update. Section 3.3 defines an intrinsic reward r^i_t = \hat{s}_t - s_t (a vector) and a combined reward R = \beta r^i_t + r^e_t (Eq. 8), and Section 4.3.2 reports that varying \beta changes performance. However, Algorithm 1 line 6 stores only the extrinsic reward r(s_t, a_t), and the Q-loss (Eq. 9), policy loss (Eq. 11), and temperature loss (Eq. 12) all operate on this stored reward with no \beta r^i_t term. Consequently the derivative of every loss with respect to \beta is identically zero, so Fig. 3c cannot be produced by the described method. Even if the reward were inserted, r^i_t is a vector difference, so Eq. 8 would add a vector to a scalar reward. The curiosity component of RMC is therefore internally inconsistent as presented.
  2. [§3.4.2, Eq. (11), Algorithm 1 line 15] The policy loss appears to be negated relative to standard SAC. Eq. (11) defines L_policy(\theta) = E[Q_{\phi_1}(s, \tilde{a}_\theta) - \alpha \log \pi_\theta(\tilde{a}_\theta|s)], and Algorithm 1 line 15 performs stochastic gradient descent on this loss. In SAC the actor is trained to maximize Q plus the entropy bonus, which corresponds to minimizing E[\alpha \log \pi - Q]. As written, gradient descent on Eq. (11) would push the policy toward minimizing Q-values and decreasing entropy relative to the stated objective, so the policy-improvement step is inverted. This is a load-bearing error in the central algorithm.
  3. [§3.4.1, Eq. (9), Algorithm 1 line 18] The Q-function target is incomplete. The target in Eq. (9) is r + (1-d)(min_{i=1,2} Q_{\phi_i}(s',\tilde{a}) - \alpha \log \pi(\tilde{a}|s')), but no discount factor \gamma multiplies the continuation term, so the Bellman backup does not match the discounted objective in Eq. (1). Additionally, Algorithm 1 line 18 updates \phi_i \leftarrow \rho \phi_i + (1-\rho)\phi_i. If \phi_i is the online parameter, this leaves it unchanged; if it is intended to denote the target parameter, the notation is identical to the online parameter, making the update ambiguous. The target-network mechanism is therefore not correctly specified.
  4. [§4.1 and §4.2] The central empirical claim that 'RMC significantly outperforms the baselines' is not backed by quantitative evidence. Section 4.1 presents only unlabeled training curves in Fig. 2 (a)-(e) with the statement that averages over five seeds were used, but no final scores, standard deviations, or statistical significance tests are reported. Similarly, the POMDP generalization results in Fig. 3a are plotted without error bars or confidence intervals. The hyperparameters for the baselines are not fully specified, and no model hyperparameters (e.g., learning rate for the model, \beta schedule, network sizes beyond the critic/actor) are given. These omissions prevent the reader from assessing the strength or reproducibility of the claimed state-of-the-art results.
  5. [§4.3.1, Fig. 3b] The choice of training scheme is justified on a single environment. Section 4.3.1 reports that 'updating RNN on model loss, value loss and not on policy loss gets the best result' based on experiments on HalfCheetah-v2 only. The paper does not show that this conclusion transfers to Ant, Humanoid, Walker2d, or the POMDP setting, yet the method's general claims rely on this scheme. This is a load-bearing gap because the value/model joint representation is the core novelty of RMC.
minor comments (6)
  1. [§3.2, Eq. (7)] Eq. (7) is labeled an L1 model loss but the formula uses the L2 norm \|\cdot\|_2. The text states that L1 outperforms L2 in the authors' experiments, but no ablation supporting this is shown. The equation should be corrected to the L1 norm or the notation clarified.
  2. [§3.1.1, after Eq. (6)] The sentence 'We will discuss the choice of back-propagating value loss and model loss later in section ??' contains an unresolved cross-reference 'section ??'. This indicates the manuscript is incomplete.
  3. [Algorithm 1, line 6] The stored tuple in line 6, (o_t, s_t, a_t, r(s_t,a_t), o_{t+1}, s_{t+1}), is inconsistent with the sampled batch in line 9, {(s,a,r,s',d)}: the terminal flag d appears only in the latter, and the observations o are not used in the loss definitions. The notation should be aligned.
  4. [§3.4.1, Eq. (9)] The phrase 'tow target Q networks' should be 'two target Q networks'. Also, the entropy term in the target is written with the current policy \pi_\theta, but in SAC it should be evaluated with the target policy; the text does not clarify this distinction.
  5. [§4.3.2, Fig. 3c] The reported effect of the curiosity scale \beta is described qualitatively ('if we use a huge scale of \beta, the intrinsic reward dominates'), but no numerical values for \beta or its decay schedule are provided, so the ablation cannot be reproduced.
  6. [§4.2, Fig. 2f] The caption of Fig. 2f says 'normalized score for training on Flicker MuJoCo with p = 0.5', and the text compares against 'standard SAC1 combine with frame stack', but the baseline configuration (e.g., number of stacked frames) is not specified. This should be stated for a fair comparison.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the paper's claims are empirical comparisons against external baselines, and no load-bearing derivation reduces to its own inputs.

full rationale

No circularity found in the derivation chain. The central claims—RMC outperforming SAC, TD3, PPO, and DDPG on MuJoCo, degrading less than SAC on flickering POMDP evaluation, and the ablation results—are empirical claims tested against external baselines and externally defined environments, not conclusions derived by definition from the method's inputs. The method sections derive standard SAC updates (Eqs. 9, 11, 12) and a model prediction loss (Eq. 7); none of these equations is fitted to the target result, and no fitted parameter is renamed as a prediction. There are no load-bearing self-citations: the references are to prior external work (SAC, R2D2, TD3, world models, L1 prediction losses), and none of the present authors' own prior results is invoked to forbid alternatives or to supply a uniqueness theorem. A real internal inconsistency exists—the curiosity reward defined in Eq. 8 as R = beta·r^i_t + r^e_t is never added to the stored reward in Algorithm 1, line 6, nor to the Q-loss (Eq. 9), policy loss (Eq. 11), or temperature loss (Eq. 12), so the curiosity ablation in Fig. 3c cannot follow from the stated algorithm as written—but this is a correctness/consistency defect, not circularity: it does not make any claimed prediction equivalent to an input by construction. Similarly, the unresolved cross-reference to 'section ??' is a missing-support defect, not a circular step. The external benchmark comparisons keep the empirical claim independently falsifiable, so the derivation chain is self-contained.

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

The central claim is empirical, so the ledger consists mainly of hyperparameters and assumptions about representation learning and evaluation. The most consequential free parameters are the curiosity weight beta, the entropy temperature alpha, and the replay/burn-in lengths. The key domain assumptions concern the sufficiency of the RNN hidden state, the joint training scheme, and the validity of the flickering benchmark. No new physical or ontological entities are introduced.

free parameters (6)
  • Curiosity reward weight beta = decayed from unspecified values to 0
    Beta scales the intrinsic curiosity reward in Eq. 8; no initial value or decay schedule is reported.
  • Entropy temperature alpha = auto-tuned via Eq. 12 with target entropy H
    Alpha trades exploration against exploitation in SAC; the target entropy H is not specified for each task.
  • Replay sequence length l_train = 15
    Used for RNN training from replay; chosen without sensitivity analysis.
  • Burn-in prefix length l_burn_in = 10
    Used to initialize the RNN hidden state for replay; chosen without sensitivity analysis.
  • Learning rate = 1e-4
    Adam learning rate applied to all networks; no per-task tuning or variation described.
  • Actor and critic network sizes = 400 and 300 hidden units
    Two-layer feed-forward networks for SAC; the RNN and model head sizes are not specified.
assumptions (6)
  • domain assumption The POMDP posterior q(s_t | o_{<=t}, a_{<t}) can be approximated by an RNN with inputs (o_t, a_{t-1}, hidden state).
    Section 3.1.1, Eq. 6. This is the core representation assumption that lets the agent use a single observation per step.
  • ad hoc to paper Blocking policy gradients into the RNN and jointly training on model and value losses yields a better representation.
    Section 3.1.1 and Section 4.3.1; supported only by the HalfCheetah-v2 ablation, not by evidence on other tasks.
  • ad hoc to paper An L1 loss on latent state deltas is better than L2 for the world model.
    Section 3.2 states L1 outperforms L2 with reference to Luo et al., but no comparison is shown in this paper.
  • domain assumption Stored-state replay with burn-in (R2D2-style) is beneficial for training the RNN.
    Section 3.1.2 relies on Kapturowski et al. 2018; adopted without an ablation here.
  • domain assumption The Flickering MuJoCo environment with observation drop probability p = 0.5 is a valid POMDP testbed.
    Section 4.2 introduces this benchmark without validating its difficulty or comparing to other POMDP benchmarks.
  • domain assumption Baseline implementations (SAC, TD3, PPO, DDPG) are properly tuned and the comparison is fair.
    Section 4.1 provides no hyperparameters or code for baselines, so the claimed advantage is unverifiable.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Reinforcement learning with world model." pith.science (2026). https://pith.science/paper/QJP54ODN

@misc{pith2026190811494,
  author       = {Pith},
  title        = {Pith review of: Reinforcement learning with world model},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QJP54ODN}},
  note         = {Machine review of arXiv:1908.11494}
}
read the original abstract

Nowadays, model-free reinforcement learning algorithms have achieved remarkable performance on many decision making and control tasks, but high sample complexity and low sample efficiency still hinder the wide use of model-free reinforcement learning algorithms. In this paper, we argue that if we intend to design an intelligent agent that learns fast and transfers well, the agent must be able to reflect key elements of intelligence, like intuition, Memory, PredictionandCuriosity. We propose an agent framework that integrates off-policy reinforcement learning with world model learning, so as to embody the important features of intelligence in our algorithm design. We adopt the state-of-art model-free reinforcement learning algorithm, Soft Actor-Critic, as the agent intuition, and world model learning through RNN to endow the agent with memory, curiosity, and the ability to predict. We show that these ideas can work collaboratively with each other and our agent (RMC) can give new state-of-art results while maintaining sample efficiency and training stability. Moreover, our agent framework can be easily extended from MDP to POMDP problems without performance loss.

Figures

Figures reproduced from arXiv: 1908.11494 by the authors.

Figure 1
Figure 1. Overall architecture of our RMC agent, which is composed of three main parts: RNN, Model head, Intuition [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 2
Figure 2. (a) to (e) are the training curves on continuous control benchmarks. RMC agent performs well consistently [PITH_FULL_IMAGE:figures/full_fig_p008_2.png] view at source ↗
Figure 3
Figure 3. (a) Agents are first trained on MDP environments and then evaluated on POMDP environments. RMC’s [PITH_FULL_IMAGE:figures/full_fig_p009_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

21 extracted references · 9 canonical work pages

  1. [1]

    Fujimoto, S., van Hoof, H., and Meger, D. (2018). Addressing function approximation error in actor-critic methods. arXiv preprint arXiv:1802.09477

  2. [2]

    P., Ghahramani, Z., Turner, R

    Gu, S., Lillicrap, T. P., Ghahramani, Z., Turner, R. E., and Levine, S. (2016). Q-prop: Sample-efficient policy gradient with an off-policy critic. ArXiv , abs/1611.02247

  3. [3]

    Guo, X. (2017). Deep Learning and Reward Design for Reinforcement Learning . PhD thesis

  4. [4]

    and Schmidhuber, J

    Ha, D. and Schmidhuber, J. (2018). World models. arXiv preprint arXiv:1803.10122

  5. [5]

    Haarnoja, T., Zhou, A., Abbeel, P., and Levine, S. (2018a). Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor. arXiv preprint arXiv:1801.01290

  6. [6]

    Haarnoja, T., Zhou, A., Hartikainen, K., Tucker, G., Ha, S., Tan, J., Kumar, V., Zhu, H., Gupta, A., Abbeel, P., et al. (2018b). Soft actor-critic algorithms and applications. arXiv preprint arXiv:1812.05905

  7. [7]

    Hafner, D., Lillicrap, T., Fischer, I., Villegas, R., Ha, D., Lee, H., and Davidson, J. (2018). Learning latent dynamics for planning from pixels. arXiv preprint arXiv:1811.04551

  8. [8]

    and Stone, P

    Hausknecht, M. and Stone, P. (2015). Deep recurrent q-learning for partially observable mdps. In 2015 AAAI Fall Symposium Series

Show all 21 references
  1. [9]

    Henderson, P., Islam, R., Bachman, P., Pineau, J., Precup, D., and Meger, D. (2018). Deep reinforcement learning that matters. In Thirty-Second AAAI Conference on Artificial Intelligence

  2. [10]

    Kapturowski, S., Ostrovski, G., Quan, J., Munos, R., and Dabney, W. (2018). Recurrent experience replay in distributed reinforcement learning

  3. [11]

    Kennedy, J. (2006). Swarm intelligence. In Handbook of nature-inspired and innovative computing , pages 187--219. Springer

  4. [12]

    Kingma, D. P. and Ba, J. (2014). Adam: A method for stochastic optimization. arXiv preprint arXiv:1412.6980

  5. [13]

    Kurutach, T., Clavera, I., Duan, Y., Tamar, A., and Abbeel, P. (2018). Model-ensemble trust-region policy optimization. arXiv preprint arXiv:1802.10592

  6. [14]

    P., Hunt, J

    Lillicrap, T. P., Hunt, J. J., Pritzel, A., Heess, N., Erez, T., Tassa, Y., Silver, D., and Wierstra, D. (2015). Continuous control with deep reinforcement learning. arXiv preprint arXiv:1509.02971

  7. [15]

    Loewenstein, G. (1994). The psychology of curiosity: A review and reinterpretation. Psychological bulletin , 116(1):75

  8. [16]

    Luo, Y., Xu, H., Li, Y., Tian, Y., Darrell, T., and Ma, T. (2018). Algorithmic framework for model-based deep reinforcement learning with theoretical guarantees. arXiv preprint arXiv:1807.03858

  9. [17]

    A., Veness, J., Bellemare, M

    Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., Graves, A., Riedmiller, M., Fidjeland, A. K., Ostrovski, G., et al. (2015). Human-level control through deep reinforcement learning. Nature , 518(7540):529

  10. [18]

    Schulman, J., Wolski, F., Dhariwal, P., Radford, A., and Klimov, O. (2017). Proximal policy optimization algorithms. arXiv preprint arXiv:1707.06347

  11. [19]

    R., Yang, C., McGreavy, C., and Li, Z

    Song, D. R., Yang, C., McGreavy, C., and Li, Z. (2018). Recurrent deterministic policy gradient method for bipedal locomotion on rough terrain challenge. In 2018 15th International Conference on Control, Automation, Robotics and Vision (ICARCV) , pages 311--318. IEEE

  12. [20]

    Todorov, E., Erez, T., and Tassa, Y. (2012). Mujoco: A physics engine for model-based control. In 2012 IEEE/RSJ International Conference on Intelligent Robots and Systems , pages 5026--5033. IEEE

  13. [21]

    Ziebart, B. D. (2010). Modeling purposeful adaptive behavior with the principle of maximum causal entropy . PhD thesis, figshare

Pith tools

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