Pith. sign in

REVIEW 5 major objections 4 minor 1 cited by

Deep Reinforcement Learning Agents are not even close to Human Intelligence

T0 review · 5 major / 4 minor · reviewed 2026-08-07 · deepseek-v4-flash

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

Pith's one-line read Deep RL agents systematically collapse on simpler versions of their training tasks, a new benchmark shows.

desk verdict HackAtari is a useful new benchmark, but the 'systematic shortcut learning' headline is stronger than the paper's own tables support. read the letter →

arxiv 2505.21731 v1 pith:UHO2HJ47 submitted 2025-05-27 cs.LG cs.AI

Quentin Delfosse , Jannis Blüml , Fabian Tatai , Théo Vincent , Bjarne Gregori , Elisabeth Dillies , Jan Peters , Constantin Rothkopf
show 1 more author
Kristian Kersting
This is my paper · ORCID
classification cs.LGcs.AI
keywords deepreinforcementlearninggeneralizationshortcutHackAtariArcadeEnvironmenttasksimplificationobject-centricRLhumanstudy
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 introduces HackAtari, a suite of over 224 RAM-level modifications of Atari 2600 games, mostly simplifications that preserve the task's goal while making it easier. It argues that deep RL agents trained on the original games collapse on these simpler variants, while non-expert humans maintain or improve their scores on most tested variations. The performance drops are read as evidence that the agents learned shortcuts tied to the training environment rather than the underlying game logic. Object-centric agents are more robust to visual changes but still fail on gameplay changes. The paper concludes that matching human scores on the training environment is not evidence of human-like reasoning, and that evaluation on held-out task variations should be standard practice.

What carries the argument

The load-bearing object is HackAtari, a benchmark of over 224 controlled variations of Arcade Learning Environment games, implemented by directly overwriting the games' Random Access Memory. RAM alteration is the mechanism because the original games are proprietary; it lets the authors remove or freeze enemies, change colors, stop cars, and alter physics in ways that preserve task semantics while reducing difficulty. The second load-bearing piece is the performance-change metric normalized against random scores, which measures an agent's relative drop from original to modified environment, together with a human study on 15 of the variations that validates the chosen variants as true simplifications.

What would settle it

Run the same trained agents on a set of RAM perturbations that are equally large in RAM space but semantically meaningless for the game, such as randomizing unused memory cells or changing the color of irrelevant background objects. If the agents' normalized scores drop by the same magnitude on these controls as on the HackAtari simplifications, then the drops would be explained by input sensitivity rather than by reliance on task-specific shortcuts, contradicting the paper's interpretation.

Watch

Extended reading notes

Core claim

The central discovery is that task simplification, not complexification, exposes misalignment in deep RL. Using RAM alterations to make games objectively easier—removing enemies, stopping cars, aligning ice floes, shifting shields by a pixel—the paper finds that the aggregated normalized score of every evaluated deep agent drops substantially, often toward random, while humans improve or hold their performance on 13 of the 15 tested variations. Because the variations were chosen to be simplifications for humans, the agents' failure is interpreted not as an inability to cope with harder tasks but as reliance on shortcuts: they succeed in the original environment for the wrong reasons. The pattern holds across value-based, policy-gradient, and object-centric agents, with object-centric representations reducing sensitivity to visual changes but not resolving gameplay-logic failures. In-distribution evaluation with human-normalized scores is therefore insufficient to certify human-like intelligence.

Load-bearing premise

The load-bearing premise is that the 15 human-tested variations are true simplifications that preserve the task's meaning, so the agents' performance drops on them indicate shortcut reliance rather than mere sensitivity to altered inputs.

Editorial extensions

If this is right

  • A superhuman score on the original environment no longer counts as evidence of task understanding, because held-out simplifications can falsify it.
  • Evaluation protocols for general RL should include both complexifications and simplifications, since only simplifications clearly separate poor adaptation from misaligned policies.
  • Object-centric inductive biases improve robustness to visual changes but are not sufficient for generalization to gameplay-logic changes.
  • Agents that pass a broad set of semantically validated simplifications are more likely to be aligned with the intended task goals.
  • Reward maximization during training alone is not enough to produce policies that generalize the way humans do.

Reading between the lines

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

  • The RAM-alteration technique could be reused to build semantic control conditions for other emulated benchmarks, separating shortcut reliance from mere input sensitivity.
  • Fine-tuning an agent on a single simplified variant and measuring how quickly it recovers could reveal whether failures come from missing features in the policy or from an inability to reuse learned skills.
  • The 224-variation suite could be distilled into a standardized, human-validated subset for reporting generalization, making agent comparisons more meaningful across studies.
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

5 major / 4 minor

Summary. This paper introduces HackAtari, a RAM-based modification framework for the Arcade Learning Environment containing over 224 variations across more than 33 games, and uses it to argue that deep RL agents rely on shortcuts. The authors evaluate six deep RL algorithms and several object-centric agents on a subset of 17 games (32 variations), reporting that agents' performance drops on simplified variants while a human study of 128 participants on 15 game-variation pairs shows that humans mostly maintain or improve performance. They conclude that training and testing in the same environment is insufficient to obtain agents with human-like generalization, and they release the benchmark and promise to release code and checkpoints. The paper includes extended results in appendices, including raw scores for all agents and human participants.

Significance. If the central claim were established, this would be a useful benchmark and a cautionary result: simplification testing is underused in RL, and the open release of 224 RAM-based variations would support robustness evaluation. The paper's strengths include the use of IQM with bootstrap confidence intervals, multiple algorithms and seeds, and a reasonably sized human study with performance-contingent bonuses. However, the headline claim of systematic performance drops is not yet supported by the aggregated evidence, and the human validation is selection-confounded. The paper is best viewed as a benchmark resource plus a promising but incomplete diagnostic study; the central conclusion requires additional representative sampling and per-variation analysis before it can be accepted.

major comments (5)
  1. [Section 4 (Q2) and Appendix F.2] The 15 human-validated variations were explicitly selected "for which most agents exhibit performance drops while we expect humans to improve or maintain their performances." Consequently, the human study confirms the selection criterion rather than validating the broader claim that HackAtari variations are simplifications. Since Figure 3 aggregates only 32 of the 224 variations and no random or held-out subset is analyzed, the paper's "systematic" conclusion is not established for the benchmark as a whole; a random sample of variations, or an explicit limitation of the claim to the selected subset, is required.
  2. [Appendix F.2 and Section 4 (human protocol)] The human protocol always presents the original game before the modified game (Free Training, Eval1 original, Eval2 modified), with no counterbalancing across participants. Human improvement in Eval2 is therefore confounded with practice or fatigue, and the paper itself invokes fatigue to explain the MsPacman drop. A counterbalanced order, or an analysis of score trajectories within each phase, is needed before human adaptation can be attributed to the simplification itself rather than to continued practice.
  3. [Table 5 and Figure 3] Table 5 contains numerous agent-variation entries with large improvements, for example C51 on Asterix obelix (7,766 to 81,812), MDQN on Bowling top pins (33 to 181), and IMPALA on StarGunner static flyers (166,881 to 878,106), and Figure 3 reports that IMPALA maintains superhuman average performance. The aggregated IQM in Figure 3 can mask these positive cases, so the abstract's claim that agents "systematically exhibit huge performance drops" requires a per-variation distribution, such as the proportion of drops and the median drop, rather than a single aggregate over games.
  4. [Figures 4 and 5, Table 7] The inference from performance drops to shortcut learning requires that the tested variations are genuine simplifications that preserve task semantics, but two of the 15 human-validated variations, BankHeist two police cars and MsPacman set level 1, produce human performance decreases or deep-agent scores near the random baseline; the text itself describes BankHeist as a decrease. These variations should be separated from strict simplifications, and the shortcut-learning interpretation should be applied only to variations with verified human non-decrease; otherwise the observed drop could reflect sensitivity to RAM-level distribution shift rather than shortcut reliance.
  5. [Appendix B and Figure 5] The object-centric comparison in Figure 5 is confounded by training budget: Appendix B states that Binary Mask PPO and Planes PPO were trained for only 40 million frames, whereas the pixel PPO baseline and the deep agents were trained for 200 million frames. Since Figure 5 uses this comparison to conclude that object-centric inductive biases are insufficient, the training budgets must be matched or explicitly accounted for in the analysis.
minor comments (4)
  1. [Section 4 and Figure 3 caption] The main text says agents were evaluated "on the 30 modifications," while the Figure 3 caption says "32 task variations," and Appendix C mentions 50 game configurations; these counts should be harmonized.
  2. [Figure 5 caption] The caption says "gameplay modifications (left)" twice; the second "left" should likely be "right," since the figure places visual perturbations on the left and gameplay modifications on the right.
  3. [Section 4 and Appendix F] The main text says "We randomly selected 134 users," while the preceding paragraph and Appendix F report 128 participants after exclusions; clarify the recruitment sample versus the analysis sample.
  4. [Appendix A and Table 7] The HNS metric is defined with human and random reference scores from Badia et al. (2020a), but Table 7 also reports in-house random scores that differ substantially (e.g., Bowling random 63.12 versus Badia random 23.1); state explicitly which random baseline is used for normalization in each figure.

Circularity Check

1 steps flagged · score 6.0 of 10

The central human-agent contrast is criterion-selected: the 15 variants used to validate 'simplification' were chosen because agents drop and humans were expected to improve, so the 'systematic' divergence is partly constructed by the selection rather than independently demonstrated.

  1. other [Section 4, Q2 ('Human adaptation to simplification far exceeds RL agents’ ones'), text preceding Figure 4; see also Section 3 definition of simplification]
    "We thus selected 15 games, with one variation for each, for which most agents exhibit performance drops while we expect humans to improve or maintain their performances."

    Section 3 defines a 'simplification' as a variation 'that will lead humans to increase or maintain their overall performances.' The Q2 validation set is then explicitly chosen as the 15 games 'for which most agents exhibit performance drops while we expect humans to improve or maintain their performances.' Thus the Figure 4 pattern—humans maintain/improve on the modified tasks while every deep agent drops—is a restatement of the inclusion criterion for those 15 variants, not an independent test of the full 224-variation HackAtari suite. The paper does not report a random or held-out sample, so the abstract's 'systematically' is constructed by selection rather than demonstrated across the suite.

full rationale

The paper's core measurement—that trained RL agents score lower on the 30 HackAtari modifications than on the originals (Figure 3)—is an independent, non-circular empirical result: the scores are obtained from frozen agents, no parameter is fitted to the modified environments, and the drop magnitudes are not derived from the training objectives. The neurosymbolic and object-centric comparisons are likewise direct evaluations. The circularity sits one layer up, in the validation of the 'simplification' label. Section 3 defines a simplification as a variation 'that will lead humans to increase or maintain their overall performances.' Section 4 then selects the human-validation set as the 15 games 'for which most agents exhibit performance drops while we expect humans to improve or maintain their performances.' The subsequent Figure 4 result—humans maintain/improve while every deep agent drops—is therefore a confirmation of the selection criterion rather than an independent characterization of HackAtari's 224 variations. The paper does not report a random or held-out sample of the suite, so the word 'systematically' in the abstract generalizes beyond what the experimental design can establish. The fixed order (free training, original evaluation, then modified evaluation) additionally makes the human improvement partly attributable to practice, though the paper notes fatigue/overfitting for MsPacman. Self-citations to Delfosse et al. (2024b) for the Pong paddle-shortcut mechanism are used as background and are not the sole load-bearing evidence; the present paper's own drop measurements carry the argument. The Section 7 limitation about absent difficulty scaling is an honest caveat but does not repair the selection issue. Overall: one central validation step is selection-circular, so the claim of a systematic human-agent divergence is only partially supported; the underlying agent-drop measurements remain legitimate.

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

The central claim rests on the interpretation of benchmark results rather than on new mathematical derivations. The main premises are that HackAtari variations are faithful simplifications, that drops indicate shortcuts, that expert scores are transferable, and that RAM edits are clean. The selection of human-study variations based on expected agent failures is the most fragile premise.

assumptions (4)
  • domain assumption The selected HackAtari variations are true simplifications that preserve task semantics.
    The paper validates this via a human study, but only for 15 variations selected in part because agents were expected to fail; MsPacman and BankHeist are acknowledged as not strict simplifications.
  • ad hoc to paper A performance drop on a simplification is evidence of shortcut learning.
    The paper interprets drops on simpler tasks as misalignment, but an alternative explanation is sensitivity to specific RAM-level distribution shifts that a correctly aligned policy could also exhibit.
  • domain assumption Expert human scores from Badia et al. (2020a) remain valid references for modified tasks.
    The HNS normalization uses expert scores from unmodified games; modified tasks may have different score ranges, though performance change uses raw scores.
  • domain assumption RAM modifications affect only the intended game attribute.
    The paper does not verify exhaustively that each RAM alteration has no unintended side effects on game logic or observations.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Deep Reinforcement Learning Agents are not even close to Human Intelligence." pith.science (2026). https://pith.science/paper/UHO2HJ47

@misc{pith2026250521731,
  author       = {Pith},
  title        = {Pith review of: Deep Reinforcement Learning Agents are not even close to Human Intelligence},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/UHO2HJ47}},
  note         = {Machine review of arXiv:2505.21731}
}
read the original abstract

Deep reinforcement learning (RL) agents achieve impressive results in a wide variety of tasks, but they lack zero-shot adaptation capabilities. While most robustness evaluations focus on tasks complexifications, for which human also struggle to maintain performances, no evaluation has been performed on tasks simplifications. To tackle this issue, we introduce HackAtari, a set of task variations of the Arcade Learning Environments. We use it to demonstrate that, contrary to humans, RL agents systematically exhibit huge performance drops on simpler versions of their training tasks, uncovering agents' consistent reliance on shortcuts. Our analysis across multiple algorithms and architectures highlights the persistent gap between RL agents and human behavioral intelligence, underscoring the need for new benchmarks and methodologies that enforce systematic generalization testing beyond static evaluation protocols. Training and testing in the same environment is not enough to obtain agents equipped with human-like intelligence.

Figures

Figures reproduced from arXiv: 2505.21731 by the authors.

Figure 2
Figure 2. Examples of HackAtari simple tasks variations. Top: the original Atari games used to trained RL agents. Bottom: simplifications (i.e. variations for which human performances do not drop). These include color changes and gameplay shifts. Superposed frames show the game dynamics. Descriptions of more environments and their variations are provided in Appendix G. alterations, such as changing or obscuring object colors,… view at source ↗
Figure 1
Figure 1. RAM alteration allows for modi￾fied environments, here exemplified on Pong. Altering specific RAM cells leads to an en￾emy remaining static after it returned the ball. This is exemplified in [PITH_FULL_IMAGE:figures/full_fig_p004_1.png] view at source ↗
Figure 3
Figure 3. Deep and symbolic RL agents performances drop on HackAtari variations, illustrated by the IQM (following reliable (Agarwal et al., 2021)) over the human normalized scores (HNS) of various RL agents on a total set of 32 task variations (over 17 games). IQMs are computed over 3 seeded trained agents (30 evaluations each). Expert-human scores are borrowed from Badia et al. (2020a). Performance in the original environme… view at source ↗
Figures from the paper (12 more)
Figure 4
Figure 4. Figure 4: While humans easily adapt to task simplifications, deep agents’ performances drop, illustrated on 15 ALE games. Non-expert users and deep RL agents are trained and evaluated on the original ALE environment, then presented with a variation of the task. Left: Variations …
Figure 5
Figure 5. Figure 5: Object-centric RL agents also fail to adapt to simplified environments. Different object-centric approaches (all using PPO) are here compared to the classical CNN baseline on the same 15 variations (as [PITH_FULL_IMAGE:figures/full_fig_p008_5.png]
Figure 3
Figure 3. Figure 3: For modified [PITH_FULL_IMAGE:figures/full_fig_p016_3.png]
Figure 6
Figure 6. Figure 6: Consent form shown to participants at the beginning of the study. It includes a brief [PITH_FULL_IMAGE:figures/full_fig_p030_6.png]
Figure 7
Figure 7. Figure 7: Payment information screen shown before the start of the task. It outlines base compen [PITH_FULL_IMAGE:figures/full_fig_p031_7.png]
Figure 8
Figure 8. Figure 8: Example game description page shown to participants before gameplay. It provides an [PITH_FULL_IMAGE:figures/full_fig_p032_8.png]
Figure 9
Figure 9. Figure 9: Pre-task questionnaire to get basic information about experience. This could support the [PITH_FULL_IMAGE:figures/full_fig_p033_9.png]
Figure 10
Figure 10. Figure 10: Screenshot of the interactive HackTari gameplay interface. Participants use this interface [PITH_FULL_IMAGE:figures/full_fig_p034_10.png]
Figure 11
Figure 11. Figure 11: Post-task questionnaire to get feedback. [PITH_FULL_IMAGE:figures/full_fig_p034_11.png]
Figure 12
Figure 12. Figure 12: Raw human scores across three phases: Free Training, Evaluation on the original task [PITH_FULL_IMAGE:figures/full_fig_p035_12.png]
Figure 13
Figure 13. Figure 13: Visualization of all results of all users. [PITH_FULL_IMAGE:figures/full_fig_p036_13.png]
Figure 14
Figure 14. Figure 14: This presentation emphasizes the robustness of human generalization to simplified [PITH_FULL_IMAGE:figures/full_fig_p037_14.png]

Discussion (0). Continue with ORCID to comment.

Forward citations

Cited by 1 Pith paper

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

  1. Bridging Physical Reasoning and Task Generalization via Visual Action Outcome Reasoning Alignment

    cs.AI 2026-07 conditional novelty 6.0 of 10

    VAORA aligns VLM chain-of-thought reasoning with visual scene observations and post-action outcomes via structured symbolic rewards, achieving cross-task and cross-environment generalization on physical reasoning benchmarks.

Reference graph

Works this paper leans on

82 extracted references · 78 canonical work pages · cited by 1 Pith paper

  1. [1]

    write newline

    " write newline "" before.all 'output.state := FUNCTION n.dashify 't := "" t empty not t #1 #1 substring "-" = t #1 #2 substring "--" = not "--" * t #2 global.max substring 't := t #1 #1 substring "-" = "-" * t #2 global.max substring 't := while if t #1 #1 substring * t #2 global.max substring 't := if while FUNCTION format.date year duplicate empty "emp...

  2. [2]

    S., Courville, A

    Agarwal, R., Schwarzer, M., Castro, P. S., Courville, A. C., and Bellemare, M. Deep reinforcement learning at the edge of the statistical precipice. Advances in Neural Information Processing Systems, 2021

    2021

  3. [3]

    R., Smith, K

    Allen, K. R., Smith, K. A., and Tenenbaum, J. B. Rapid trial-and-error learning with simulation supports flexible tool use and physical reasoning. Proceedings of the National Academy of Sciences, 2020

    work page 2020

  4. [4]

    The option-critic architecture

    Bacon, P., Harb, J., and Precup, D. The option-critic architecture. In Proceedings of the Thirty-First AAAI Conference on Artificial Intelligence , 2017

    work page 2017

  5. [5]

    P., Piot, B., Kapturowski, S., Sprechmann, P., Vitvitskyi, A., Guo, Z

    Badia, A. P., Piot, B., Kapturowski, S., Sprechmann, P., Vitvitskyi, A., Guo, Z. D., and Blundell, C. Agent57: Outperforming the atari human benchmark. In Proceedings of the 37th International Conference on Machine Learning,, 2020 a

    work page 2020

  6. [6]

    P., Sprechmann, P., Vitvitskyi, A., Guo, Z

    Badia, A. P., Sprechmann, P., Vitvitskyi, A., Guo, Z. D., Piot, B., Kapturowski, S., Tieleman, O., Arjovsky, M., Pritzel, A., Bolt, A., and Blundell, C. Never give up: Learning directed exploration strategies. In 8th International Conference on Learning Representations, 2020 b

    work page 2020

  7. [7]

    L., Saxe, R., and Tenenbaum, J

    Baker, C. L., Saxe, R., and Tenenbaum, J. B. Action understanding as inverse planning. Cognition, 2009

    work page 2009

  8. [8]

    H., Wang, R., and Manchester, I

    Barbara, N. H., Wang, R., and Manchester, I. On robust reinforcement learning with lipschitz-bounded policy networks. In ICML Workshop: Foundations of Reinforcement Learning and Control--Connections and Perspectives, 2024

    work page 2024

Show all 82 references
  1. [9]

    Verifiable reinforcement learning via policy extraction

    Bastani, O., Pu, Y., and Solar-Lezama, A. Verifiable reinforcement learning via policy extraction. In Advances in Neural Information Processing Systems, 2018

    2018

  2. [10]

    W., Hamrick, J

    Battaglia, P. W., Hamrick, J. B., and Tenenbaum, J. B. Simulation as an engine of physical scene understanding. Proceedings of the National Academy of Sciences, 2013

    2013

  3. [11]

    G., Naddaf, Y., Veness, J., and Bowling, M

    Bellemare, M. G., Naddaf, Y., Veness, J., and Bowling, M. The arcade learning environment: An evaluation platform for general agents. Journal of Artificial Intelligence Research, 2013

    2013

  4. [12]

    G., Dabney, W., and Munos, R

    Bellemare, M. G., Dabney, W., and Munos, R. A distributional perspective on reinforcement learning. In Proceedings of the 34th International Conference on Machine Learning, 2017

    2017

  5. [13]

    Multi-objective causal bayesian optimization

    Bhatija, S., Zuercher, P.-D., Thumm, J., and Bohn \'e , T. Multi-objective causal bayesian optimization. arXiv, 2025

    2025

  6. [14]

    Deep reinforcement learning via object-centric attention

    Bl \"u ml, J., Derstroff, C., Gregori, B., Dillies, E., Delfosse, Q., and Kersting, K. Deep reinforcement learning via object-centric attention. arXiv, 2025

    2025

  7. [15]

    Cannon, T. P. and S im s ek, \"O . Accelerating task generalisation with multi-level hierarchical options. In The Thirteenth International Conference on Learning Representations, 2025

    2025

  8. [16]

    Galois: boosting deep reinforcement learning via generalizable logic synthesis

    Cao, Y., Li, Z., Yang, T., Zhang, H., Zheng, Y., Li, Y., Hao, J., and Liu, Y. Galois: boosting deep reinforcement learning via generalizable logic synthesis. Advances in Neural Information Processing Systems, 2022

    2022

  9. [17]

    Chan, S. C. Y., Fishman, S., Korattikara, A., Canny, J. F., and Guadarrama, S. Measuring the reliability of reinforcement learning algorithms. In 8th International Conference on Learning Representations, 2020

    2020

  10. [18]

    Quantifying generalization in reinforcement learning

    Cobbe, K., Klimov, O., Hesse, C., Kim, T., and Schulman, J. Quantifying generalization in reinforcement learning. In Proceedings of the 36th International Conference on Machine Learning, 2019

    2019

  11. [19]

    Leveraging procedural generation to benchmark reinforcement learning

    Cobbe, K., Hesse, C., Hilton, J., and Schulman, J. Leveraging procedural generation to benchmark reinforcement learning. In International conference on machine learning, 2020

    2020

  12. [20]

    and Lake, B

    Davidson, G. and Lake, B. M. Investigating simple object representations in model-free deep reinforcement learning. In Proceedings of the 42th Annual Meeting of the Cognitive Science Society, 2020

    2020

  13. [21]

    D., Gershman, S

    Daw, N. D., Gershman, S. J., Seymour, B., Dayan, P., and Dolan, R. J. Model-based influences on humans' choices and striatal prediction errors. Neuron, 2011

    2011

  14. [22]

    S., and Kersting, K

    Delfosse, Q., Shindo, H., Dhami, D. S., and Kersting, K. Interpretable and explainable logical policies via neurally guided symbolic abstraction. Advances in Neural Information Processing (NeurIPS), 2023 a

    2023

  15. [23]

    Boosting object representation learning via motion and object continuity

    Delfosse, Q., Stammer, W., Rothenbacher, T., Vittal, D., and Kersting, K. Boosting object representation learning via motion and object continuity. In European Conference on Machine Learning and Principles and Practice of Knowledge Discovery in Databases ( ECML ) , 2023 b

    2023

  16. [24]

    OCAtari : O bject-centric Atari 2600 reinforcement learning environments

    Delfosse, Q., Bl \" u ml, J., Gregori, B., Sztwiertnia, S., and Kersting, K. OCAtari : O bject-centric Atari 2600 reinforcement learning environments. Reinforcement Learning Journal, 2024 a

    2024

  17. [25]

    Interpretable concept bottlenecks to align reinforcement learning agents

    Delfosse, Q., Sztwiertnia, S., Stammer, W., Rothermel, M., and Kersting, K. Interpretable concept bottlenecks to align reinforcement learning agents. Advances in Neural Information Processing Systems, 2024 b

    2024

  18. [26]

    Dennett, D. C. The intentional stance. 1989

    1989

  19. [27]

    L., Koch, J., Sharkey, L

    di Langosco, L. L., Koch, J., Sharkey, L. D., Pfau, J., and Krueger, D. Goal misgeneralization in deep reinforcement learning. In International Conference on Machine Learning, 2022

    2022

  20. [28]

    P., and Kersting, K

    Dillies, E., Delfosse, Q., Bl \"u ml, J., Emunds, R., Busch, F. P., and Kersting, K. Better decisions through the right causal world model. arXiv, 2025

    2025

  21. [29]

    and Rothkopf, C

    Dimitrakakis, C. and Rothkopf, C. A. Bayesian multitask inverse reinforcement learning. In European workshop on reinforcement learning, 2011

    2011

  22. [30]

    Diuk, C., Cohen, A., and Littman, M. L. An object-oriented representation for efficient reinforcement learning. In Proceedings of the 25th international conference on Machine learning, 2008

    2008

  23. [31]

    IMPALA: scalable distributed deep-rl with importance weighted actor-learner architectures

    Espeholt, L., Soyer, H., Munos, R., Simonyan, K., Mnih, V., Ward, T., Doron, Y., Firoiu, V., Harley, T., Dunning, I., Legg, S., and Kavukcuoglu, K. IMPALA: scalable distributed deep-rl with importance weighted actor-learner architectures. In Proceedings of the 35th Internation...

    2018

  24. [32]

    C., and Bowling, M

    Farebrother, J., Machado, M. C., and Bowling, M. Generalization and regularization in DQN . 2018

    2018

  25. [33]

    Forbus, K. D. Qualitative physics: Past, present, and future. In Exploring artificial intelligence. 1988

    1988

  26. [34]

    S., Brendel, W., Bethge, M., and Wichmann, F

    Geirhos, R., Jacobsen, J., Michaelis, C., Zemel, R. S., Brendel, W., Bethge, M., and Wichmann, F. A. Shortcut learning in deep neural networks. Nat. Mach. Intell., 2020

    2020

  27. [35]

    Atari agents, 2022

    Gogianu, F., Berariu, T., Bușoniu, L., and Burceanu, E. Atari agents, 2022

    2022

  28. [36]

    Visualizing and understanding atari agents, 2018

    Greydanus, S., Koul, A., Dodge, J., and Fern, A. Visualizing and understanding atari agents, 2018

    2018

  29. [37]

    E., Pechenizkiy, M., and Mocanu, D

    Grooten, B., Sokar, G., Dohare, S., Mocanu, E., Taylor, M. E., Pechenizkiy, M., and Mocanu, D. C. Automatic noise filtering with dynamic sparse training in deep reinforcement learning. In International Conference on Autonomous Agents and Multiagent Systems, 2023

    2023

  30. [38]

    H., Codel, C., Hofmann, K., Houghton, B., Kuno, N., Milani, S., Mohanty, S., Liebana, D

    Guss, W. H., Codel, C., Hofmann, K., Houghton, B., Kuno, N., Milani, S., Mohanty, S., Liebana, D. P., Salakhutdinov, R., Topin, N., et al. The minerl competition on sample efficient reinforcement learning using human priors. arXiv, 2019

    2019

  31. [39]

    Benchmarking the spectrum of agent capabilities

    Hafner, D. Benchmarking the spectrum of agent capabilities. In International Conference on Learning Representations, 2022

    2022

  32. [40]

    P., Ba, J., and Norouzi, M

    Hafner, D., Lillicrap, T. P., Ba, J., and Norouzi, M. Dream to control: Learning behaviors by latent imagination. In 8th International Conference on Learning Representations, 2020

    2020

  33. [41]

    T., Wang, Z., Heess, N., and Riedmiller, M

    Hausman, K., Springenberg, J. T., Wang, Z., Heess, N., and Riedmiller, M. Learning an embedding space for transferable robot skills. In International Conference on Learning Representations, 2018

    2018

  34. [42]

    S., and Kersting, K

    Helff, L., Stammer, W., Shindo, H., Dhami, D. S., and Kersting, K. V-lol: A diagnostic dataset for visual logical learning. Journal of Data-centric Machine Learning Research, 2025

    2025

  35. [43]

    Rainbow: Combining improvements in deep reinforcement learning

    Hessel, M., Modayil, J., Van Hasselt, H., Schaul, T., Ostrovski, G., Dabney, W., Horgan, D., Piot, B., Azar, M., and Silver, D. Rainbow: Combining improvements in deep reinforcement learning. In Proceedings of the AAAI conference on artificial intelligence, 2018

    2018

  36. [44]

    Huang, S., Dossa, R. F. J., Ye, C., Braga, J., Chakraborty, D., Mehta, K., and Araújo, J. G. Cleanrl: High-quality single-file implementations of deep reinforcement learning algorithms. Journal of Machine Learning Research, 2022

    2022

  37. [45]

    Adversarial examples are not bugs, they are features

    Ilyas, A., Santurkar, S., Tsipras, D., Engstrom, L., Tran, B., and Madry, A. Adversarial examples are not bugs, they are features. Advances in neural information processing systems, 2019

    2019

  38. [46]

    and Luo, S

    Jiang, Z. and Luo, S. Neural logic reinforcement learning. In International Conference on Machine Learning, 2019

    2019

  39. [47]

    u ml, J., W \

    Kaufmann, T., Bl \"u ml, J., W \"u st, A., Delfosse, Q., Kersting, K., and H \"u llermeier, E. Ocalm: Object-centric assessment with language models. arXiv, 2024

    2024

  40. [48]

    Kessler, F., Frankenstein, J., and Rothkopf, C. A. Human navigation strategies and their errors result from dynamic interactions of spatial uncertainties. Nature Communications, 2024

    2024

  41. [49]

    Objective robustness in deep reinforcement learning

    Koch, J., Langosco, L., Pfau, J., Le, J., and Sharkey, L. Objective robustness in deep reinforcement learning. arXiv, 2021

    2021

  42. [50]

    Interpretable and editable programmatic tree policies for reinforcement learning

    Kohler, H., Delfosse, Q., Akrour, R., Kersting, K., and Preux, P. Interpretable and editable programmatic tree policies for reinforcement learning. arXiv, 2024

    2024

  43. [51]

    R., Holyoak, K

    Kubricht, J. R., Holyoak, K. J., and Lu, H. Intuitive physics: Current research and controversies. Trends in cognitive sciences, 2017

    2017

  44. [52]

    M., Ullman, T

    Lake, B. M., Ullman, T. D., Tenenbaum, J. B., and Gershman, S. J. Building machines that learn and think like people. Behavioral and brain sciences, 2017

    2017

  45. [53]

    Sub-policy adaptation for hierarchical reinforcement learning

    Li, A., Florensa, C., Clavera, I., and Abbeel, P. Sub-policy adaptation for hierarchical reinforcement learning. In International Conference on Learning Representations, 2020

    2020

  46. [54]

    V., Sun, W., Singh, G., Deng, F., Jiang, J., and Ahn, S

    Lin, Z., Wu, Y., Peri, S. V., Sun, W., Singh, G., Deng, F., Jiang, J., and Ahn, S. SPACE: unsupervised object-oriented scene representation via spatial attention and decomposition. In International Conference on Learning Representations, 2020

    2020

  47. [55]

    Hierarchical programmatic reinforcement learning via learning to compose programs

    Liu, G.-T., Hu, E.-P., Cheng, P.-J., Lee, H.-Y., and Sun, S.-H. Hierarchical programmatic reinforcement learning via learning to compose programs. In International Conference on Machine Learning, 2023

    2023

  48. [56]

    Insight: End-to-end neuro-symbolic visual reinforcement learning with language explanations

    Luo, L., Zhang, G., Xu, H., Yang, Y., Fang, C., and Li, Q. Insight: End-to-end neuro-symbolic visual reinforcement learning with language explanations. arXiv, 2024

    2024

  49. [57]

    C., Bellemare, M

    Machado, M. C., Bellemare, M. G., Talvitie, E., Veness, J., Hausknecht, M., and Bowling, M. Revisiting the arcade learning environment: Evaluation protocols and open problems for general agents. Journal of Artificial Intelligence Research, 2018

    2018

  50. [58]

    Sympol: Symbolic tree-based on-policy reinforcement learning

    Marton, S., Grams, T., Vogt, F., L \"u dtke, S., Bartelt, C., and Stuckenschmidt, H. Sympol: Symbolic tree-based on-policy reinforcement learning. In International Conference on Learning Representations, 2025

    2025

  51. [59]

    The perception of causality

    Michotte, A. The perception of causality. 1963

    1963

  52. [60]

    H., Mohanty, S

    Milani, S., Topin, N., Houghton, B., Guss, W. H., Mohanty, S. P., Nakata, K., Vinyals, O., and Kuno, N. S. Retrospective analysis of the 2019 minerl competition on sample efficient reinforcement learning. In NeurIPS 2019 competition and demonstration track, 2020

    2019

  53. [61]

    A., Veness, J., Bellemare, M

    Mnih, V., Kavukcuoglu, K., Silver, D., Rusu, A. A., Veness, J., Bellemare, M. G., Graves, A., Riedmiller, M. A., Fidjeland, A., Ostrovski, G., Petersen, S., Beattie, C., Sadik, A., Antonoglou, I., King, H., Kumaran, D., Wierstra, D., Legg, S., and Hassabis, D. Human-level cont...

    2015

  54. [62]

    Robust reinforcement learning: A review of foundations and recent advances

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

    2022

  55. [63]

    Neup \"a rtl, N., Tatai, F., and Rothkopf, C. A. Intuitive physical reasoning about objects’ masses transfers to a visuomotor decision task consistent with newtonian physics. PLoS Computational Biology, 2020

    2020

  56. [64]

    Robust adversarial reinforcement learning

    Pinto, L., Davidson, J., Sukthankar, R., and Gupta, A. Robust adversarial reinforcement learning. In International conference on machine learning, 2017

    2017

  57. [65]

    Proximal policy optimization algorithms

    Schulman, J., Wolski, F., Dhariwal, P., Radford, A., and Klimov, O. Proximal policy optimization algorithms. 2017

    2017

  58. [66]

    Schultz, W., Dayan, P., and Montague, P. R. A neural substrate of prediction and reward. Science, 1997

    1997

  59. [67]

    Reinforcement learning in strategy-based and atari games: A review of google deepminds innovations

    Shaheen, A., Badr, A., Abohendy, A., Alsaadawy, H., and Alsayad, N. Reinforcement learning in strategy-based and atari games: A review of google deepminds innovations. arXiv, 2025

    2025

  60. [68]

    S., and Kersting, K

    Shindo, H., Delfosse, Q., Dhami, D. S., and Kersting, K. Blendrl: A framework for merging symbolic and neural policy learning. In International Conference on Learning Representations, 2025

    2025

  61. [69]

    Silver, D., Singh, S., Precup, D., and Sutton, R. S. Reward is enough. Artificial intelligence, 2021

    2021

  62. [70]

    Right for the right concept: Revising neuro-symbolic concepts by interacting with their explanations

    Stammer, W., Schramowski, P., and Kersting, K. Right for the right concept: Revising neuro-symbolic concepts by interacting with their explanations. In Proceedings of the IEEE/CVF conference on computer vision and pattern recognition, 2021

    2021

  63. [71]

    Straub, D., Schultheis, M., Koeppl, H., and Rothkopf, C. A. Probabilistic inverse optimal control for non-linear partially observable systems disentangles perceptual uncertainty and behavioral costs. Advances in Neural Information Processing Systems, 2023

    2023

  64. [72]

    Scaling up robust mdps using function approximation

    Tamar, A., Mannor, S., and Xu, H. Scaling up robust mdps using function approximation. In Proceedings of the 31st International Conference on Machine Learning, 2014

    2014

  65. [73]

    Tatai, F., Straub, D., and Rothkopf, C. A. Intuitive sensorimotor decisions under risk take newtonian physics into account. 2025

    2025

  66. [74]

    B., Kemp, C., Griffiths, T

    Tenenbaum, J. B., Kemp, C., Griffiths, T. L., and Goodman, N. D. How to grow a mind: Statistics, structure, and abstraction. Science, 2011

    2011

  67. [75]

    Domain randomization for transferring deep neural networks from simulation to the real world

    Tobin, J., Fong, R., Ray, A., Schneider, J., Zaremba, W., and Abbeel, P. Domain randomization for transferring deep neural networks from simulation to the real world. In international conference on intelligent robots and systems, 2017

    2017

  68. [76]

    Munchausen reinforcement learning

    Vieillard, N., Pietquin, O., and Geist, M. Munchausen reinforcement learning. Advances in Neural Information Processing Systems, 2020

    2020

  69. [77]

    Iterated q -network: Beyond one-step bellman updates in deep reinforcement learning

    Vincent, T., Palenicek, D., Belousov, B., Peters, J., and D'Eramo, C. Iterated q -network: Beyond one-step bellman updates in deep reinforcement learning. Transactions on Machine Learning Research, 2025

    2025

  70. [78]

    S., Rothkopf, C

    W \"u st, A., Tobiasch, T., Helff, L., Ibs, I., Stammer, W., Dhami, D. S., Rothkopf, C. A., and Kersting, K. Bongard in wonderland: Visual puzzles that still make ai go mad? arXiv, 2025

    2025

  71. [79]

    Towards generalizable reinforcement learning via causality-guided self-adaptive representations

    Yang, Y., Huang, B., Feng, F., Wang, X., Tu, S., and Xu, L. Towards generalizable reinforcement learning via causality-guided self-adaptive representations. In International Conference on Learning Representations, 2025

    2025

  72. [80]

    F., Raposo, D., Santoro, A., Bapst, V., Li, Y., Babuschkin, I., Tuyls, K., Reichert, D

    Zambaldi, V. F., Raposo, D., Santoro, A., Bapst, V., Li, Y., Babuschkin, I., Tuyls, K., Reichert, D. P., Lillicrap, T. P., Lockhart, E., Shanahan, M., Langston, V., Pascanu, R., Botvinick, M. M., Vinyals, O., and Battaglia, P. W. Relational deep reinforcement learning. arXiv, 2018

    2018

  73. [81]

    Natural environment benchmarks for reinforcement learning

    Zhang, A., Wu, Y., and Pineau, J. Natural environment benchmarks for reinforcement learning. arXiv, 2018

    2018

  74. [82]

    N., et al

    Zhu, Y., Gao, T., Fan, L., Huang, S., Edmonds, M., Liu, H., Gao, F., Zhang, C., Qi, S., Wu, Y. N., et al. Dark, beyond deep: A paradigm shift to cognitive ai with humanlike common sense. Engineering, 2020

    2020

Pith tools

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