Pith. sign in

REVIEW 3 major objections 5 minor 13 references

Maximum Total Correlation Reinforcement Learning

T0 review · 3 major / 5 minor · reviewed 2026-08-07 · deepseek-v4-flash

Pith's one-line read Total-correlation maximization makes RL trajectories simpler and more resilient to perturbations.

desk verdict Solid empirical regularizer, but the title's total-correlation mechanism is not what the optimized objective does; needs reframing and a few consistency fixes. read the letter →

arxiv 2505.16734 v1 pith:P3HND7KT submitted 2025-05-22 cs.LG

classification cs.LG
keywords totalcorrelationreinforcementlearningtrajectorycompressibilityrobustcontrolvariationallowerboundsoftactor-critictemporalconsistencysimplicitybias
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 tries to establish that adding an auxiliary objective to reinforcement learning (maximize the total correlation among the states and actions inside a trajectory) is a task-independent way to induce simpler behavior and, with it, better generalization. Total correlation measures how much information is saved by encoding an entire trajectory jointly instead of compressing each time step separately, so maximizing it pushes policies toward periodic, predictable, open-loop-like behavior. The authors derive a variational lower bound on this trajectory-level objective, fold it into soft actor-critic as a regularized reward, and report that the resulting agent (MTC) out-scores SAC, RPC, LZ-SAC, and SPAC on most DeepMind Control tasks, matches or exceeds them on Metaworld and image-based control, and degrades less under observation noise, action noise, and body-mass changes. If the claims hold, a single extra information-theoretic term could serve as a general robustness regularizer for continuous control.

What carries the argument

The central object is the total correlation of a trajectory, $\mathcal{C}(z_1; a_1; \ldots; a_{T-1}; z_T) = \mathbb{E}[\log p(z_1, a_1, \ldots, a_{T-1}, z_T) / \prod_t p(z_t) \prod_t p(a_t)]$, which quantifies how many nats are saved by encoding the whole trajectory rather than each state and action separately. The argument is carried by the variational lower bound in Eq. (2), which replaces the unknown marginals with the learned encoder $f_\theta(z_{t+1}|s_{t+1})$ and policy $\pi_\phi(a_t|s_t)$ and adds a history-based dynamics model $q_\eta(z_{t+1}|z_{1:t}, a_{1:t})$ and action prediction model $q_\chi(a_t|z_{1:t}, a_{1:t-1})$. This bound turns total-correlation maximization into a stepwise, differentiable reward bonus that favors states and actions whose future is predictable from the past, and it is what allows the objective to be optimized with a standard actor-critic loop.

What would settle it

Train the same agent with the total-correlation reward replaced by a constant negative offset of matched magnitude: if compressibility and perturbation robustness are unchanged, the bound's internal predictive structure is not what matters. Alternatively, estimate the true total correlation of rollout trajectories from joint and marginal entropies before and after MTC training; if true total correlation decreases while the lower bound increases, the paper's proposed mechanism is refuted.

Watch

Extended reading notes

Core claim

On its own terms, the paper's central claim is that trajectories induced by a policy can be made simpler and more robust by maximizing their total correlation, defined as the KL divergence between the joint distribution of the sequence of latent states and actions and the product of their per-time-step marginals. Because the exact quantity is intractable, the paper optimizes a variational lower bound built from a history-based latent dynamics model and a history-based action prediction model; the bound rewards situations in which the next latent state and the next action are predictable from the past, relative to the irreducible randomness of the encoder and policy. The resulting algorithm, MTC, is soft actor-critic with an extra reward term plus a dual constraint that automatically tunes the strength of the total-correlation term. Empirically, the paper reports that MTC's trajectories are the most compressible in bzip2 file-size comparisons, that its actions are the most predictable to a t-step-ahead predictor, and that it maintains higher returns than baselines under Gaussian observation noise, action noise, mass-scaling perturbations, and added irrelevant state dimensions.

Load-bearing premise

The load-bearing premise is that maximizing a bound that is guaranteed to be negative, and therefore cannot estimate total correlation at all, still shifts policies toward genuinely higher total correlation and produces the reported robustness gains.

Editorial extensions

If this is right

  • If the central claim is right, a single trajectory-level information-theoretic term can improve both task performance and zero-shot robustness on locomotion, manipulation, and image-based control benchmarks.
  • MTC-learned policies should produce measurably more compressible trajectories and more predictable future actions; the paper reports both (bzip2 sizes and t-step-ahead prediction errors).
  • The regularizer should generalize across tasks without per-task reward engineering, because it acts only on the induced trajectory distribution, not on task-specific structure.
  • The approach can be layered onto any off-policy actor-critic method that can accept a modified reward; the paper demonstrates this on soft actor-critic and reports similar gains on DMC, Metaworld, and image-based DMC tasks.
  • The action-prediction component specifically carries part of the robustness benefit; the paper's ablation MTC-NoA indicates that removing it reduces robustness under several perturbations.

Reading between the lines

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

  • Editorial inference: since the optimized lower bound is always negative, the practical objective is a weighted sum of KL divergences between predictive models and policy/encoder distributions; if that is the true driver, similar gains might come from any information-bottleneck-style regularizer, not specifically from total correlation.
  • Editorial inference: the robustness gains are reported at fixed noise and mass scales; a sharper test would be to check whether MTC's advantage grows with trajectory length and with the amount of redundant periodic structure in the task, which would confirm the compressibility mechanism.
  • Editorial inference: combining MTC with domain randomization could reduce the amount of simulator variation needed for sim-to-real transfer, since the consistency bias already removes some dependence on state-level details; the paper does not test this combination.
Share X Bluesky LinkedIn Reddit HN

Signed reviews

No signed human review yet.

Editorial analysis

A structured set of objections, weighed in public.

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

Referee Report

3 major / 5 minor

Summary. The paper proposes Maximum Total Correlation Reinforcement Learning (MTC-RL), a modification of the RL objective that adds a total-correlation term over the induced trajectory of latent states and actions, justified as an inductive bias toward simple, compressible, and robust behavior. The authors derive a variational lower bound (Eq. 2) using history-based latent dynamics and action prediction models, integrate it into SAC-style off-policy optimization with an adaptive Lagrangian coefficient, and evaluate the method on DMC, Metaworld, and image-based DMC benchmarks. They report improved asymptotic performance, enhanced zero-shot robustness to observation/action noise and dynamics changes, higher trajectory compressibility, and stronger action predictability compared to SAC, RPC, LZ-SAC, and SPAC. The paper also includes a discussion section acknowledging that the lower bound is always negative and hence not useful for estimating the true total correlation.

Significance. If the central claim were established, MTC-RL would provide a principled, task-independent simplicity bias for RL with broad empirical appeal. The experimental evaluation is substantial and careful: 20 seeds on DMC, multiple benchmarks, ablations, and public code. The variational derivation in Appendix A.1 is mathematically correct as a lower bound, and the robustness and compressibility improvements are reproducible in principle. However, the theoretical interpretation is not supported: the optimized surrogate (Eq. 2) is a sum of negated KL divergences and is always non-positive, so maximizing it is not equivalent to maximizing total correlation. The paper explicitly concedes this in Section 6. The empirical gains may be real, but they are currently attributed to a mechanism that the optimized objective does not implement. This mismatch is load-bearing for the title, abstract, and interpretation of every empirical result. The paper therefore requires major revision, either by providing evidence that optimizing the surrogate increases true total correlation or by reframing the contribution as conditional-KL regularization.

major comments (3)
  1. [§4.2, Eq. (2); §6] The variational lower bound eC in Eq. (2) is always non-positive because it is a sum of negated KL divergences. The authors acknowledge this in Section 6, but the implication for the paper's central claim is stronger than "not useful for estimation": maximizing eC does not correspond to maximizing total correlation. Specifically, eC = -Σ_t E[KL(fθ(z_{t+1}|s_{t+1}) || qη(z_{t+1}|history)) + KL(πφ(a_t|s_t) || qχ(a_t|history))]. A policy that ignores the state and outputs independent actions drawn from the marginal, together with q matching the encoder and policy marginals, attains eC = 0, the theoretical maximum of the bound, while the true total correlation of the trajectory is near zero. Conversely, a deterministic periodic trajectory can have large true total correlation while eC is near zero because both fθ and qη are sharply peaked. Thus the bound's gradients push toward conditional independence of each variable from the current state given the history, not toward the multi-variable dependence measured by total correlation. The title and abstract claim that the method "maximizes the total correlation within the induced trajectories"; this claim is not established. To support it, the paper should measure the true total correlation on a tractable task (e.g., a low-dimensional MDP) and show that it increases when optimizing Eq. (2), or it should explicitly reframe the method as state-conditional KL regularization.
  2. [§5.3, Fig. 3] The trajectory-compression result in Section 5.3 and Fig. 3 is presented as evidence that the method maximizes total correlation. However, compressibility is a direct consequence of the optimized objective: Eq. (2) explicitly minimizes the KL divergence between the encoder/policy and history-based predictive models, which by construction makes trajectories predictable and hence compressible. This is not an independent confirmation of the total-correlation mechanism; the same compression improvement would be expected from any regularizer that penalizes KL to a history-based predictor, including a simple action-smoothing baseline. To make the empirical claim non-circular, the paper should compare against a baseline that optimizes the same KL regularizer without the total-correlation interpretation, and should additionally report an independent estimate of total correlation (e.g., a non-parametric plug-in estimate) on the same trajectories.
  3. [§4.2, Eq. (2); Appendix A.1] The claim in §4.2 that the lower bound "captures important aspects of the total correlation" is substantiated only by a qualitative trajectory plot (Fig. 1), which does not establish a quantitative relationship between eC and the true total correlation. The gap between the true C and eC is D_KL(p(z_{1:T},a_{1:T-1}) || q(z_{1:T},a_{1:T-1})) plus additional non-negative terms from the marginal-substitution step in Eq. (11). This gap can be large and input-dependent, so the optimization of eC need not track the optimization of C. The paper should provide at least one empirical or analytic example where eC and C move in opposite directions, or a small-scale measurement showing that increasing eC indeed increases C. Without this, the central mechanism remains unsupported.
minor comments (5)
  1. [§5.4 vs. Fig. 4 caption] The text in §5.4 states that MTC "achieves higher average rewards than baselines on all tasks" for Metaworld, while the caption of Fig. 4 says "MTC is competitive to baselines"; these statements should be reconciled.
  2. [Table 3, Appendix B.9] The entry "Replay buffer capacity 1 00 000" contains a typo; it should read "100 000".
  3. [Eq. (3)] Equation (3) writes r(s_T, a_T) separately in front of the sum over t=1 to T-1, which is redundant with the t=T term inside the sum and may confuse the indexing; please clean up the notation.
  4. [§6] The statement that a vacuous lower bound can be useful for optimization "as in the case of subtracting a constant offset" is imprecise: a lower bound with an input-dependent gap is not the same as a constant offset, and maximizing a lower bound is only justified if the gap is controlled or if the bound is tight at the optimum. This point should be clarified or removed.
  5. [Table 1] Table 1 reports "means over 20 seeds with 90% confidence interval" but the values are shown as ± intervals; please clarify whether these are standard errors, standard deviations, or quantile-based intervals.

Circularity Check

0 steps flagged · score 0.0 of 10

No circularity: the lower-bound derivation is genuine (though vacuous), and the admitted always-negative surrogate is a soundness gap, not a self-referential input-output identity.

full rationale

The paper's derivation chain is self-contained: Eq. 1 defines the total-correlation-augmented objective, Eq. 2 is a variational lower bound on that total correlation (Appendix A.1), and Eq. 3 optimizes the bound with SAC. The bound is mathematically valid, so the algorithm is not defined in terms of the conclusion. Section 6 explicitly discloses the serious limitation: 'our lower bound of the total correlation corresponds to a sum of negated KL divergences, and is therefore always negative. Hence, it is not useful for estimating the actual total correlation.' Because eC <= 0 and total correlation C >= 0, the inequality C >= eC is trivially true, so the derivation does not establish that maximizing eC maximizes C. This is a correctness/soundness gap in the total-correlation interpretation, not a circular reduction: the paper does not silently rename the surrogate as the target, and the robustness/compressibility results are zero-shot empirical evaluations against external baselines rather than fitted quantities renamed as predictions. No load-bearing self-citation appears: co-author citations (Peters et al. 2010; Memmel et al. 2022; Reddi et al. 2023) are contextual related work, and the RPC/LZ-SAC discussion relies on external prior work. The finding is therefore a non-finding for circularity, with the vacuous-bound issue recorded here as the main correctness risk.

Assumptions & free parameters 3 free parameters · 5 assumptions · 0 invented entities

The central derivation relies on standard variational bounds, but the load-bearing proxy assumption is that the vacuous lower bound can be optimized to increase true total correlation. The main free parameters are per-task tuned Ip, history length, and bound-combination weight m. No new physical or conceptual entities are introduced.

free parameters (3)
  • Information constraint Ip = -0.5, -7.0, -10.0, -0.1 per task
    Controls the target lower bound for the total correlation objective. Tuned per task via hyperparameter search, as stated in Appendix B.3 and Tables 2-3.
  • History length = 8 for state-based tasks, 3 or 5 for image-based tasks
    Number of past steps fed to the LSTM-based prediction models. Tuned per task and reported in Tables 2-3.
  • Bound combination coefficient m = 1e-6 for state-based, 1e-4 for image-based
    Weight of the action bound in the combined lower bound. Tuned separately for state-based and image-based settings.
assumptions (5)
  • standard math The variational lower bound in Eq. 2 is a valid lower bound on total correlation
    Follows from non-negativity of KL divergence; derivation in Appendix A.1.
  • standard math Substituting the encoder conditional fθ(zt+1|st+1) and policy conditional πϕ(at|st) for the unknown marginals p(zt+1) and p(at) preserves a lower bound
    Relies on non-negativity of the added KL terms in Eq. 11.
  • ad hoc to paper Maximizing the always-negative lower bound is a useful proxy for maximizing total correlation
    The bound is vacuous as an estimator, as the authors state in Section 6. The paper assumes it is still a valid optimization target without proof.
  • domain assumption Optimizing the prediction models on the replay buffer (off-policy) instead of on-policy samples does not invalidate the bound
    Stated in Section 4.4 as a deviation that 'may increase the gap of our lower bound'; the paper proceeds without analyzing the impact.
  • domain assumption A history-free policy using only the current state is sufficient despite the history-dependent reward
    The main experiments use a Markov policy; the paper provides an ablation in Appendix C.5 showing similar performance on Finger Spin, but this is only tested on one task.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Maximum Total Correlation Reinforcement Learning." pith.science (2026). https://pith.science/paper/P3HND7KT

@misc{pith2026250516734,
  author       = {Pith},
  title        = {Pith review of: Maximum Total Correlation Reinforcement Learning},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/P3HND7KT}},
  note         = {Machine review of arXiv:2505.16734}
}
read the original abstract

Simplicity is a powerful inductive bias. In reinforcement learning, regularization is used for simpler policies, data augmentation for simpler representations, and sparse reward functions for simpler objectives, all that, with the underlying motivation to increase generalizability and robustness by focusing on the essentials. Supplementary to these techniques, we investigate how to promote simple behavior throughout the episode. To that end, we introduce a modification of the reinforcement learning problem that additionally maximizes the total correlation within the induced trajectories. We propose a practical algorithm that optimizes all models, including policy and state representation, based on a lower-bound approximation. In simulated robot environments, our method naturally generates policies that induce periodic and compressible trajectories, and that exhibit superior robustness to noise and changes in dynamics compared to baseline methods, while also improving performance in the original tasks.

Figures

Figures reproduced from arXiv: 2505.16734 by the authors.

Figure 1
Figure 1. Maximizing the total correlation within trajectories results in more consistent behavior. As shown in our experiments, this consistency increases robustness to noise and dynamics changes. Abstract Simplicity is a powerful inductive bias. In re￾inforcement learning, regularization is used for simpler policies, data augmentation for simpler representations, and sparse reward functions for simpler objectives, all that,… view at source ↗
Figure 2
Figure 2. We evaluated the robustness towards observation noise (left), action noise (middle) and mass changes (right) on eight tasks from DMC benchmarks. The plots show the normalized mean rewards averaged over 20 independent runs and 8 tasks, with error bars representing 90% confidence interval. For each task we normalized the return by the mean return of the best method. Each run includes 30 evaluation trajectories. MTC ac… view at source ↗
Figure 3
Figure 3. The compressed state-action trajectories obtained by MTC have smallest file size in expectation. Based on these visualizations, we argue that MTC produces the simplest and most consistent trajectories, characterized by highly cyclical patterns. To support this qualitative assessment, inspired by Saanum et al. (2023), we use lossless compression algorithms to quantify the compressibility of trajectories produced by l… view at source ↗
Figures from the paper (10 more)
Figure 4
Figure 4. Figure 4: Left: aggregated performance of MTC and baselines at 500K environment steps on six image-based DMC tasks. The plot shows the normalized average rewards over 5 runs and 6 tasks, with error bars representing 90% confidence interval. For each run, we collect 10 evaluation…
Figure 5
Figure 5. Figure 5: We test the robustness of MTC and its two ablations, MTC-NoA and SAC, on the Walker Stand task and the Cheetah Run task. Overall, MTC achieves better or at least comparable average rewards in the presence of observation noise (left column), action noise (middle column)…
Figure 6
Figure 6. Figure 6: We evaluate our method on eight image-based DMC tasks. We evaluate the performance of our method on the commonly used PlaNet benchmark, which consists of a set of complex image-based continuous control tasks. Specifially, we consider six tasks:Ball-in-cup Catch, Cartpo…
Figure 7
Figure 7. Figure 7: Learning curves of our method and baselines on 8 DMC tasks. The plot shows the average rewards over 20 seeds with a shading of 90% confidence interval. MTC achieves better or at least comparable performance and sample efficiency than baselines. C.2. Predictability of P…
Figure 8
Figure 8. Figure 8: Performance of our method and baselines on 5 manipulation tasks from Metaworld. The plot shows the average success rate and 90% confidence interval over 10 seeds. MTC achieves comparable performance than baselines. C.4. Image-based DMC Tasks Table. 6 shows the performa…
Figure 9
Figure 9. Figure 9: We test the performance of MTC with different constraints Ip. The robustness to mass changes and observation noise, and the compression of behaviors are improved while increasing Ip. C.7. Robustness Comparison on A Single Task A constant increase in noise could result …
Figure 10
Figure 10. Figure 10: Robustness to observation noise (left), action noise (middle), mass changes (right) on Walker Stand. The plot shows the average reward over 20 seeds, with error bars representing 90% confidence interval. MTC achieves higher rewards than baselines in the presence of st…
Figure 11
Figure 11. Figure 11: Robustness to spurious correlations on Walker Stand. MTC achieves higher rewards than RPC and SAC, when states are expanded with unrelated Gaussian noises. 0 50 100 150 200 0.0 0.5 1.0 State 0 MTC 0 50 100 150 200 0.0 0.5 RPC 0 50 100 150 200 0.5 1.0 LZ-SAC 0 50 100 1…
Figure 12
Figure 12. Figure 12: Visualizations of state sequences generated by our method and baselines on the Finger Spin task. State sequences of our method show more repeating and periodic patterns. 22 [PITH_FULL_IMAGE:figures/full_fig_p022_12.png]
Figure 13
Figure 13. Figure 13: Visualizations of action sequences generated by our method and baselines on the Finger Spin task. MTC produces more consistent and periodic behavior than baselines. 23 [PITH_FULL_IMAGE:figures/full_fig_p023_13.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

13 extracted references · 9 canonical work pages

  1. [2]

    For other details, please refer to the provided code. B.4. Extended Description of Baseline Implementations SAC. We obtain the results for SAC by running the PyTorch implementations of SAC (Yarats et al., 2021). We use the same hyperparameters for SAC as our algorithm to ensure a fair comparison. We found that our obtained results for SAC are stronger tha...

  2. [3]

    Hyperparameters used in Image-based tasks. Parameter Value Ip for Finger and Ball-In-Cup -10.0 Ip for other tasks -0.1 history length for Finger and Ball-In-Cup 3 history length for other tasks 5 Replay buffer capacity 1 00 000 Optimizer Adam Critic learning rate 10−4 Critic Q-function soft-update rate 0.01 Critic target update frequency 2 Actor learning ...

  3. [7]

    Tassa, Y ., Doron, Y ., Muldal, A., Erez, T., Li, Y ., Casas, D. d. L., Budden, D., Abdolmaleki, A., Merel, J., Lefrancq, A., et al. Deepmind control suite. arXiv preprint arXiv:1801.00690,

  4. [12]

    log p(z1:T |s1:T ) p(z1:T ) # − E

    does not always hold as it was derived by replacing p(z1:T |s1:T ) by QT −1 t=1 p(zt+1|st+1) (Eysenbach et al., 2021, Appendix C1). These distributions are in general not the same because information about future state observations can decrease uncertainty about the current latent state, and therefore p(zt+1|z1:t, s1:T ) ̸= p(zt+1|st+1). We will now show ...

  5. [13]

    Hyperparameters used in MTC. Parameter Value Ip for Cheetah Run, Hopper and Walker Stand -0.5 Ip for other tasks -7.0 history length 8 Replay buffer capacity 1 000 000 Optimizer Adam Critic learning rate 10−4 Critic Q-function soft-update rate 0.01 Critic target update frequency 2 Actor learning rate 10−4 Actor update frequency 1 Actor log stddev bounds [...

  6. [14]

    State sequences of our method show more repeating and periodic patterns

    Visualizations of state sequences generated by our method and baselines on the Finger Spin task. State sequences of our method show more repeating and periodic patterns. 22 Maximum Total Correlation Reinforcement Learning 0 25 50 75 100 125 150 175 200 1 0 1 Action Actuator 0 0 25 50 75 100 125 150 175 200 1 0 1 Actuator 1 0 25 50 75 100 125 150 175 200 1...

  7. [1934]

    URL http://www.jstor.org/ stable/jj.8441753.32

    ISBN 9780520364257. URL http://www.jstor.org/ stable/jj.8441753.32. Oord, A. v. d., Li, Y ., and Vinyals, O. Representation learn- ing with contrastive predictive coding. arXiv preprint arXiv:1807.03748,

  8. [1995]

    STEERING: Stein Information Directed Exploration for Model-Based Reinforcement Learning

    Chakraborty, S., Bedi, A. S., Koppel, A., Wang, M., Huang, F., and Manocha, D. Steering: Stein information di- rected exploration for model-based reinforcement learn- ing. arXiv preprint arXiv:2301.12038,

Show all 13 references
  1. [2010]

    and Zaslavsky, N

    Tishby, N. and Zaslavsky, N. Deep learning and the infor- mation bottleneck principle. In 2015 IEEE Information Theory Workshop (ITW), pp. 1–5. IEEE,

  2. [2018]

    for more descriptions of tasks. B.2. Implementation Details SAC codebase. We implement our algorithm on top of the common PyTorch implementation of the SAC algorithm (Yarats et al., 2021). We used the default hyperparameters from that implementation unless specified otherwise....

  3. [2021]

    Dynamics generalization via information bottleneck in deep rein- forcement learning

    Lu, X., Lee, K., Abbeel, P., and Tiomkin, S. Dynamics generalization via information bottleneck in deep rein- forcement learning. arXiv preprint arXiv:2008.00614,

  4. [2022]

    Soft actor-critic algorithms and applications

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

  5. [2024]

    doi: 10.1126/scirobotics. adi9579. URL https://www.science.org/doi/ abs/10.1126/scirobotics.adi9579. Rakelly, K., Gupta, A., Florensa, C., and Levine, S. Which mutual-information representation learning objectives are sufficient for control? Advances in Neural Information Proc...

Pith tools

Reviewed August 7, 2026 · model on record in the stance chip above.