Pith. sign in

REVIEW 4 major objections 5 minor 45 references

xSRL: Safety-Aware Explainable Reinforcement Learning -- Safety as a Product of Explainability

T0 review · 4 major / 5 minor · reviewed 2026-08-11 · deepseek-v4-flash

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

Pith's one-line read The paper claims that combining local risk critics with global policy graphs makes RL safety actionable, allowing developers to attack, debug, and patch policies without retraining.

desk verdict A useful integration of local and global explanations for RL safety with a serious typo-level flaw in the risk-critic loss that must be fixed before the central claims can be trusted. read the letter →

arxiv 2412.19311 v1 pith:7ZVER5GZ submitted 2024-12-26 cs.AI cs.HCcs.LGcs.MA

Risal Shahriar Shefin , Md Asifur Rahman , Thai Le , Sarra Alqahtani This is my paper · ORCID
classification cs.AIcs.HCcs.LGcs.MA
keywords explainablereinforcementlearningsafelocalexplanationsglobaladversarialattackpolicypatchingriskcriticgraphs
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 claims that safety in reinforcement learning can be improved as a product of explainability: if operators can see both the agent's overall strategy and its per-state risk estimates, they can identify, attack, and patch vulnerabilities without retraining the policy. xSRL trains two post-hoc critics, one for task reward ($Q_{\mathrm{task}}$) and one for safety cost ($Q_{\mathrm{risk}}$), and attaches their values to abstract-state policy graphs built by an existing global explanation method. This combination is what the authors call the first framework to join local and global explanations for RL safety and to offer adversarial explanations for vulnerability analysis and policy patching. The authors support the claim with fidelity measurements, user studies, and an attack-then-patch demonstration in two safe-RL benchmark environments.

What carries the argument

The load-bearing mechanism is the post-hoc risk critic $Q_{\mathrm{risk}}$, trained by minimizing an MSE Bellman-like loss (Eq.~2) over sampled transitions of a fixed target policy, and its task counterpart $Q_{\mathrm{task}}$. These critics are averaged over abstract states in a CAPS-style global policy graph (Eqs.~4--5), letting each node carry expected return and expected safety cost. The mechanism also includes the threshold definitions: a state is safety-critical when $Q_{\mathrm{risk}}(s,a) > \epsilon_{\mathrm{safety}}$ for some policy action (Eq.~6), and the shield triggers when $Q_{\mathrm{risk}}(s_t,a_t) > T_{\mathrm{safety}}$ (Eq.~7), replacing the action with one from a separately trained safety policy. This machinery turns explanation values into attack targets and patch triggers.

What would settle it

A reader could run the SAC agent in Navigation 2 with the same 50% attack, but with the $Q_{\mathrm{risk}}$ critic deliberately corrupted by noise or by training it on trajectories from a different policy, and check whether the set of safety-critical states and the shield's firing behavior change materially; if they do, the safety explanations inherit the critic's error. A cleaner quantitative check is to compare $Q_{\mathrm{risk}}$'s predicted cumulative cost against the ground-truth cost in held-out episodes under adversarial perturbations, and ask whether its NRMSE (reported between 33% and 63%) is concentrated precisely in the high-risk states used for attack and patching.

Watch

Extended reading notes

Core claim

On the paper's own terms, the central discovery is that a single explanation artifact—a directed graph of abstract states with $Q_{\mathrm{task}}$ and $Q_{\mathrm{risk}}$ attached, and edges labelled as safety- or task-driven—gives users both comprehension and debugging utility. With $Q_{\mathrm{risk}}$ estimated from sampled transitions for any fixed policy, safety-critical states can be defined by thresholding $Q_{\mathrm{risk}}$ (Eq.~6), attacked adversarially by forcing alternative actions at those states, and then patched by a shield that substitutes a safety-policy action when $Q_{\mathrm{risk}}$ exceeds a threshold (Eq.~7). The evaluation claims that xSRL explanations achieve the best user accuracy for safe agents (77.4%), that an explanation-guided attack on a SAC agent reduces its safety by roughly 72% at a 50% attack rate, and that patching with a separate safety policy and shield restores safety without retraining. The authors conclude that separate-policy safe RL methods produce more interpretable safety behaviour than joint-optimization methods, and that safety-critical decisions become explainable exactly because the risk critic is available.

Load-bearing premise

The whole pipeline rests on the learned risk critic $Q_{\mathrm{risk}}$ being an accurate estimate of expected safety cost for the target policy even under adversarial perturbations and in states the critic was not trained on; if that estimate is wrong, the safety-critical states, the attack targeting, and the shield decisions are all wrong.

Editorial extensions

If this is right

  • xSRL gives users a testable map of where an RL agent is unsafe, so safety audits reduce to reading high-$Q_{\mathrm{risk}}$ nodes before deployment.
  • With the same $Q_{\mathrm{risk}}$ critic used as both explanation and shield trigger, the cost of adding safety explanations is amortized into actual safety enforcement.
  • Explanation-guided attacks become a debugging workflow: contrast the graph before and after attack, identify rising $Q_{\mathrm{risk}}$ values, patch the policy, and re-explain to verify the fix.
  • Separate-policy safe RL methods (such as AdvExRL) are more explainable than joint-optimization ones (such as SQRL), so interpretability can act as a selection criterion for safe RL algorithms.
  • Users can identify the safer of two agents under attack with high confidence using the combined graph, supporting its use in operator training and incident review.

Reading between the lines

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

  • Because the two-critic recipe only assumes a discounted-sum objective, the same graph could carry critics for uncertainty, fairness, or human-specified costs, turning xSRL into a general lens for what an agent worries about.
  • The shield inherits the risk critic's estimation error, so tightening $Q_{\mathrm{risk}}$ with off-policy evaluation methods should improve explanation fidelity and shield reliability together; that is a testable prediction the paper does not make.
  • The explain-attack-patch-re-explain loop points toward a certification protocol: a policy is plausibly 'explanation-certified safe' when no explanation-guided attack at high-$Q_{\mathrm{risk}}$ states measurably degrades safety, which could serve as a benchmark for safe RL releases.
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

4 major / 5 minor

Summary. The paper proposes xSRL, a framework for safety-aware explainable reinforcement learning that combines local explanations (learned Q_task and Q_risk critics) with global explanations (an extension of the CAPS policy-graph summarization) for agents in constrained MDPs. xSRL additionally provides adversarial explanations by using Q_risk to identify safety-critical states, launching adversarial attacks at those states, and patching the policy with a safety shield or a separately trained safety policy. The evaluation comprises computational fidelity measures (action, policy-selection, Q_risk, Q_task) on SAC, AdvExRL, RRL-MF, and SQRL agents in two MuJoCo environments, plus three user studies with 270 participants measuring comprehension and ability to identify the safer agent. The authors claim that safety can be improved as a byproduct of explainability, and that xSRL is the first framework to combine local and global explanations for RL safety and to offer adversarial explanations for vulnerability analysis and policy patching.

Significance. If the results hold, xSRL would be a practically useful contribution to explainable and safe RL, integrating several desiderata (local explanations, global summaries, adversarial vulnerability analysis, and post-hoc patching without retraining) into one toolkit. The paper ships code and includes a large user study, which are strengths. However, the central claims rest on the accuracy of the learned risk critic and on fidelity numbers that are, in their reported form, partly at odds with the paper's own abstract and introduction. The contribution is novel enough for AAMAS, but the evidence needs to be tightened before the safety-as-a-product-of-explainability claim is fully supported.

major comments (4)
  1. [§5.1.1, Table 1] The risk-critic loss J_risk is written as the MSE between the current Q_risk estimate and the expected future Q_risk estimate only, omitting the immediate cost c_t and the discount factor gamma. As written, any constant Q is a minimizer, so Q_risk would not encode expected safety costs as defined in Eq. (1). This is load-bearing because Q_risk is used to define safety-critical states in Eq. (6), to trigger the shield in Eq. (7), and to generate the local and global explanations. Please correct Eq. (2) to include c_t and gamma, or supply code or an appendix demonstrating that the implemented loss actually contains these terms. The Table 1 NRMSE values (33.2% to 62.8%) do not resolve the issue by themselves, since a constant predictor can yield comparable NRMSE depending on the scale of the costs.
  2. [§5.1.1, Table 1] The fidelity results for patched agents are weak relative to the claims made in the abstract and Section 1. Action fidelity for AdvExRL, RRL-MF, and SQRL is 34.25%, 31.7%, and 32.4% respectively, and Q_risk NRMSE is 33.2%, 48.4%, and 62.8%. The introduction states that xSRL generates accurate policy graphs with less than 33.5% error in risk estimation, which is contradicted by the RRL-MF and SQRL rows. These numbers do not convincingly establish that xSRL provides high-fidelity explanations for the patched agents that are central to the safety claims. Please report error bars or significance tests across the five graphs, and revise the accuracy claims to match the actual ranges.
  3. [§5.1.2, Table 2] The user study results for H1 only partially support the claim that xSRL is the most effective explanation method. For unsafe agents, xSRL and global explanations achieve exactly the same average accuracy (60.37%), so xSRL is not superior to CAPS in the high-risk setting that motivated the paper. For safe agents, xSRL leads by 77.4% vs 73.7%, but no statistical test is reported, so it is unclear whether this difference is meaningful. The conclusion that xSRL is most effective and the related superiority claim over global explanations need to be either supported with inferential statistics or appropriately softened.
  4. [§5.2.1, Figure 2] The utility evaluation of patching is weakened by how the attack is constructed. The explanation-guided attack in Section 4.2 targets high-risk abstract states identified from the SAC agent's own Q_risk values, and the same attack is then applied to all patched agents. Since AdvExRL, RRL-MF, and SQRL already have their own safety mechanisms, the results in Figure 2 partly reflect the pre-existing robustness of those methods rather than the benefit of xSRL's patching. Moreover, there is no comparison against a baseline that patches without xSRL's explanation guidance, so the incremental utility of xSRL in the patching loop is not established.
minor comments (5)
  1. [§1] The introduction states that fidelity tests show 'less than 33.5% error in risk estimation', but Table 1 reports Q_risk NRMSE of 48.4% for RRL-MF and 62.8% for SQRL. Please correct this inconsistency.
  2. [§4.1, Eq. (1), Eq. (3)] The notation in Eqs. (1) and (3) is confusing: the expectation subscript is taken over a_t while the sum runs over future timesteps, and the inserted 'task' text in Eq. (3) is misplaced. Please reformulate these expressions cleanly.
  3. [§5.1.1, Table 1] No measures of variance are provided for the fidelity scores, despite the text stating that scores are averaged across five generated graphs. Reporting standard deviations or confidence intervals would substantially strengthen the empirical claims.
  4. [§5.1.2, User Studies] The recruitment description says '270 participants (30 per study)' with three studies, which is arithmetically inconsistent; presumably the intended meaning is 30 participants per condition across nine conditions. Please clarify.
  5. [§1] There is a typo: 'trustworthines' should be 'trustworthiness'.

Circularity Check

1 steps flagged · score 6.0 of 10

The risk-critic training loss (Eq. 2) defines Q_risk in terms of Q_risk alone, omitting the immediate cost and discount, making the core risk prediction self-referential by construction.

  1. self definitional [Section 4.1, Eq. (2), Local Explanation Method]
    "In practice, we approximate Q̂^{πtask}_{φ,risk}, parameterized by φ, using sampled transitions (s_t,a_t,s_{t+1},c_t). This is done by minimizing the following MSE loss with respect to the target (RHS of Eq. 1): J_{risk}(s_t,a_t,s_{t+1};φ) = 1/2 ( Q̂^{πtask}_{φ,risk}(s_t,a_t) − E_{a_{t+1}∼π(.|s_{t+1})}[ Q̂^{πtask}_{φ,risk}(s_{t+1},a_{t+1})] )^2 (2)"

    The text says the loss target is the RHS of Eq. 1, which contains the immediate safety cost c(s_t,a_t) and the discount γ. As written, Eq. 2's target is only the expected next-state Q_risk value, with no cost and no discount. Minimizing this loss imposes Q_risk(s,a) = E[Q_risk(s',a')] over policy transitions, a homogeneous Bellman equation whose solutions are constants (or harmonic functions) independent of the actual costs c_t. Consequently, Q_risk is defined in terms of itself rather than in terms of the safety-cost input. The downstream safety-critical-state definition (Eq. 6) and the shield trigger (Eq.

full rationale

The paper's broader framework is evaluated with external safety metrics (Safety%, Success-Safety%) and user studies, and the self-citations to CAPS and AdvExRL are disclosed, published components/baselines rather than load-bearing uniqueness claims; those do not by themselves make the work circular. However, the local risk critic Q_risk is the load-bearing mechanism: it is used to generate local explanations, to select attack targets (Eq. 6), and to trigger the proposed safety shield (Eq. 7). Eq. 2 as written omits the immediate cost c_t and the discount γ, leaving a target that is just the expected next-state Q_risk. That makes the risk estimate a self-consistency equation with constant solutions, so the claimed risk prediction is not derived from safety costs but from itself. This is a specific, quotable reduction (Eq. 2's target is Q_risk(s',a') only), not a vague impression. The remaining pipeline—graph construction, user comprehension, and external safety evaluation—has independent content, so the circularity is partial rather than total.

Assumptions & free parameters 2 free parameters · 4 assumptions · 0 invented entities

No new physical entities are introduced. The framework relies on learned Q-critics and two hand-chosen thresholds. The axioms are standard CMDP assumptions, post-hoc estimability of Q-functions, and the validity of user-study comprehension measures.

free parameters (2)
  • Tsafety = not specified (environment-specific)
    Shield trigger threshold in Eq. 7, chosen based on a sensitivity test for each environment (Section 4.3).
  • epsilon_safety = not specified
    Threshold in Definition 1 (Eq. 6) to declare a state safety-critical; no value or selection procedure given.
assumptions (4)
  • domain assumption CMDP with episodic termination upon task completion or safety violation
    Section 3 states this assumption to define the problem; it is standard but restricts applicability to episodic tasks with terminating safety violations.
  • domain assumption Post-hoc Q-critics trained on sampled transitions approximate true Q-values for the fixed target policy
    Eqs. 1-3 assume the learned critics generalize; Table 1 shows high NRMSE for some agents, limiting this assumption.
  • domain assumption The pre-trained adversarial policy from [26] (authors' own work) provides realistic alternative actions for AAA attacks
    Section 4.2 uses pi_adv from AdvExRL without independent validation; the attack's relevance depends on this policy's quality.
  • domain assumption Participant answers in user studies reliably measure comprehension of the explanations
    Section 5.1.2 uses accuracy on four questions as a proxy for trust and understanding; this is standard in XAI but unvalidated here.

how reviews work

0 comments
Cite this review

Pith. "Pith review of xSRL: Safety-Aware Explainable Reinforcement Learning -- Safety as a Product of Explainability." pith.science (2026). https://pith.science/paper/7ZVER5GZ

@misc{pith2026241219311,
  author       = {Pith},
  title        = {Pith review of: xSRL: Safety-Aware Explainable Reinforcement Learning -- Safety as a Product of Explainability},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/7ZVER5GZ}},
  note         = {Machine review of arXiv:2412.19311}
}
read the original abstract

Reinforcement learning (RL) has shown great promise in simulated environments, such as games, where failures have minimal consequences. However, the deployment of RL agents in real-world systems such as autonomous vehicles, robotics, UAVs, and medical devices demands a higher level of safety and transparency, particularly when facing adversarial threats. Safe RL algorithms have been developed to address these concerns by optimizing both task performance and safety constraints. However, errors are inevitable, and when they occur, it is essential that the RL agents can also explain their actions to human operators. This makes trust in the safety mechanisms of RL systems crucial for effective deployment. Explainability plays a key role in building this trust by providing clear, actionable insights into the agent's decision-making process, ensuring that safety-critical decisions are well understood. While machine learning (ML) has seen significant advances in interpretability and visualization, explainability methods for RL remain limited. Current tools fail to address the dynamic, sequential nature of RL and its needs to balance task performance with safety constraints over time. The re-purposing of traditional ML methods, such as saliency maps, is inadequate for safety-critical RL applications where mistakes can result in severe consequences. To bridge this gap, we propose xSRL, a framework that integrates both local and global explanations to provide a comprehensive understanding of RL agents' behavior. xSRL also enables developers to identify policy vulnerabilities through adversarial attacks, offering tools to debug and patch agents without retraining. Our experiments and user studies demonstrate xSRL's effectiveness in increasing safety in RL systems, making them more reliable and trustworthy for real-world deployment. Code is available at https://github.com/risal-shefin/xSRL.

Figures

Figures reproduced from arXiv: 2412.19311 by the authors.

Figure 1
Figure 1. Examples of generated explanations for Navigation2 task using our local explanation, global explanation from CAPS [PITH_FULL_IMAGE:figures/full_fig_p006_1.png] view at source ↗
Figure 2
Figure 2. Safety(%) and success-safety(%) performance of the [PITH_FULL_IMAGE:figures/full_fig_p007_2.png] view at source ↗
Figure 3
Figure 3. An example of using (a) the xSRL explanation graph to launch an attack on the SAC agent at high-risk states, (b) the [PITH_FULL_IMAGE:figures/full_fig_p008_3.png] view at source ↗

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

45 extracted references · 26 canonical work pages

  1. [2]

    Mohammed Alshiekh, Roderick Bloem, Rüdiger Ehlers, Bettina Könighofer, Scott Niekum, and Ufuk Topcu. 2018. Safe reinforcement learning via shielding. In Proceedings of the AAAI Conference on Artificial Intelligence , Vol. 32

    work page 2018

  2. [3]

    Dan Amir and Ofra Amir. 2018. HIGHLIGHTS: Summarizing Agent Behavior to People. In Proceedings of the 17th International Conference on Autonomous Agents and MultiAgent Systems (Stockholm, Sweden) (AAMAS ’18). International Founda- tion for Autonomous Agents and Multiagent Systems, Richland, SC, 1168–1176

    work page 2018

  3. [4]

    Ofra Amir, Finale Doshi-Velez, and David Sarne. 2019. Summarizing agent strategies. Autonomous Agents and Multi-Agent Systems33, 5 (Sept. 2019), 628–644. https://doi.org/10.1007/s10458-019-09418-w

    work page doi:10.1007/s10458-019-09418-w 2019

  4. [5]

    Osbert Bastani. 2021. Safe reinforcement learning with nonlinear dynamics via model predictive shielding. In 2021 American Control Conference (ACC) . IEEE, 3488–3494

    work page 2021

  5. [6]

    Brittany Davis, Maria Glenski, William Sealy, and Dustin Arendt. 2020. Measure Utility, Gain Trust: Practical Advice for XAI Researchers. In 2020 IEEE Workshop on TRust and EXpertise in Visual Analytics (TREX) . 1–8. https://doi.org/10.1109/ TREX51495.2020.00005

    work page arXiv 2020

  6. [7]

    Peter Geibel. 2006. Reinforcement learning for MDPs with constraints. In Euro- pean Conference on Machine Learning . Springer, 646–653

    work page 2006

  7. [8]

    Piyush Gupta, Nikaash Puri, Sukriti Verma, Dhruv Kayastha, Shripad Deshmukh, Balaji Krishnamurthy, and Sameer Singh. [n.d.]. Explain Your Move: Under- standing Agent Actions Using Specific and Relevant Feature Attribution. ([n. d.]). https://par.nsf.gov/biblio/10166401

    work page arXiv

  8. [9]

    Tuomas Haarnoja, Aurick Zhou, Pieter Abbeel, and Sergey Levine. 2018. Soft actor-critic: Off-policy maximum entropy deep reinforcement learning with a stochastic actor. In International conference on machine learning . PMLR, 1861– 1870

    2018

Show all 45 references
  1. [10]

    Bradley Hayes and J. Shah. 2017. Improving Robot Controller Transparency Through Autonomous Policy Explanation. 2017 12th ACM/IEEE International Conference on Human-Robot Interaction (HRI (2017), 303–312

    2017

  2. [11]

    Jacob Hilton, Nick Cammarata, Shan Carter, Gabriel Goh, and Christopher Olah

  3. [12]

    Fred Hohman, Minsuk Kahng, Robert Pienta, and Duen Horng Chau. 2019. Visual Analytics in Deep Learning: An Interrogative Survey for the Next Frontiers.IEEE Transactions on Visualization and Computer Graphics 25, 8 (2019), 2674–2693. https://doi.org/10.1109/TVCG.2018.2843369

    2019

  4. [13]

    Huang, David Held, Pieter Abbeel, and Anca D

    Sandy H. Huang, David Held, Pieter Abbeel, and Anca D. Dragan. 2019. Enabling robots to communicate their objectives. Auton. Robots 43, 2 (feb 2019), 309–326. https://doi.org/10.1007/s10514-018-9771-0

    2019 doi

  5. [14]

    Tobias Huber, Benedikt Limmer, and Elisabeth Andr’e. 2021. Benchmarking Perturbation-Based Saliency Maps for Explaining Atari Agents. Frontiers in Arti- ficial Intelligence 5 (2021). https://api.semanticscholar.org/CorpusID:235490210

    2021

  6. [15]

    Zoe Juozapaitis, Anurag Koul, Alan Fern, Martin Erwig, and Finale Doshi-Velez

  7. [16]

    Youngmin Kim, Richard Allmendinger, and Manuel López-Ibáñez. 2020. Safe learning and optimization techniques: Towards a survey of the state of the art. In International Workshop on the Foundations of Trustworthy AI Integrating Learning, Optimization and Reasoning. Springer, 123–139

    2020

  8. [17]

    Isaac Lage, Daphna Lifschitz, Finale Doshi-Velez, and Ofra Amir. 2019. Exploring computational user models for agent policy summarization. In Proceedings of the 28th International Joint Conference on Artificial Intelligence (Macao, China) (IJCAI’19). AAAI Press, 1401–1407

    2019

  9. [18]

    Sergey Levine, Peter Pastor, Alex Krizhevsky, Julian Ibarz, and Deirdre Quillen

  10. [19]

    Bing Liu, Yiyuan Xia, and Philip S. Yu. 2004. Clustering Via Decision Tree Construction

    2004

  11. [20]

    Aniek Markus, Jan Kors, and Peter Rijnbeek. 2021. The role of explainability in creating trustworthy artificial intelligence for health care: A comprehensive survey of the terminology, design choices, and evaluation strategies. Journal of Biomedical Informatics 113 (Jan 2021),...

    2021 doi

  12. [21]

    Joe McCalmon, Thai Le, Sarra Alqahtani, and Dongwon Lee. 2022. CAPS: Com- prehensible Abstract Policy Summaries for Explaining Reinforcement Learning Agents. In Proceedings of the 21st International Conference on Autonomous Agents and Multiagent Systems (Virtual Event, New Zea...

    2022

  13. [22]

    Oliver Mihatsch and Ralph Neuneier. 2002. Risk-sensitive reinforcement learning. Machine learning 49, 2 (2002), 267–290

    2002

  14. [23]

    Riedmiller

    Volodymyr Mnih, Koray Kavukcuoglu, David Silver, Alex Graves, Ioannis Antonoglou, Daan Wierstra, and Martin A. Riedmiller. 2013. Playing Atari with Deep Reinforcement Learning. CoRR abs/1312.5602 (2013). arXiv:1312.5602 http://arxiv.org/abs/1312.5602

    2013 arXiv

  15. [24]

    Sina Mohseni, Niloofar Zarei, and Eric D. Ragan. 2018. A Survey of Evaluation Methods and Measures for Interpretable Machine Learning.ArXiv abs/1811.11839 (2018). https://api.semanticscholar.org/CorpusID:54087635

    2018 arXiv

  16. [25]

    Christoph Molnar. 2022. Interpretable Machine Learning (2 ed.). https:// christophm.github.io/interpretable-ml-book

    2022

  17. [26]

    Md Asifur Rahman, Tongtong Liu, and Sarra Alqahtani. 2023. Adversarial Behav- ior Exclusion for Safe Reinforcement Learning. InProceedings of the Thirty-Second International Joint Conference on Artificial Intelligence, IJCAI-23 , Edith Elkind (Ed.). International Joint Confere...

    2023 doi

  18. [27]

    Arvind Satyanarayan, Dominik Moritz, Kanit Wongsuphasawat, and Jeffrey Heer

  19. [28]

    Jordan, and Pieter Abbeel

    John Schulman, Sergey Levine, Philipp Moritz, Michael I. Jordan, and Pieter Abbeel. 2015. Trust Region Policy Optimization. CoRR abs/1502.05477 (2015). arXiv:1502.05477 http://arxiv.org/abs/1502.05477

    2015 arXiv

  20. [29]

    John Schulman, Filip Wolski, Prafulla Dhariwal, Alec Radford, and Oleg Klimov

  21. [30]

    Karen Simonyan, Andrea Vedaldi, and Andrew Zisserman. 2013. Deep Inside Convolutional Networks: Visualising Image Classification Models and Saliency Maps. CoRR abs/1312.6034 (2013). https://api.semanticscholar.org/CorpusID: 1450294

    2013 arXiv

  22. [31]

    Krishnan Srinivasan, Benjamin Eysenbach, Sehoon Ha, Jie Tan, and Chelsea Finn. 2020. Learning to be safe: Deep rl with a safety critic. arXiv preprint arXiv:2010.14603 (2020)

    2020 arXiv

  23. [32]

    Chen Tessler, Yonathan Efroni, and Shie Mannor. 2019. Action robust reinforce- ment learning and applications in continuous control. InInternational Conference on Machine Learning. PMLR, 6215–6224

    2019

  24. [33]

    Brijen Thananjeyan, Ashwin Balakrishna, Suraj Nair, Michael Luo, Krishnan Srinivasan, Minho Hwang, Joseph E Gonzalez, Julian Ibarz, Chelsea Finn, and Ken Goldberg. 2021. Recovery rl: Safe reinforcement learning with learned recovery zones. IEEE Robotics and Automation Letters ...

    2021

  25. [34]

    ArXiv abs/1707.06347 (2017)

    Proximal Policy Optimization Algorithms. ArXiv abs/1707.06347 (2017). https://api.semanticscholar.org/CorpusID:28695052

    2017 arXiv

  26. [35]

    Nicholay Topin and Manuela Veloso. 2019. Generation of Policy-Level Expla- nations for Reinforcement Learning. In The Thirty-Third AAAI Conference on Artificial Intelligence, AAAI 2019, The Thirty-First Innovative Applications of Artifi- cial Intelligence Conference, IAAI 2019...

    2019

  27. [36]

    Laurens van der Maaten and Geoffrey Hinton. 2008. Visualizing Data using t-SNE. Journal of Machine Learning Research 9, 86 (2008), 2579–2605. http: //jmlr.org/papers/v9/vandermaaten08a.html

    2008

  28. [37]

    Jasper van der Waa, Jurriaan van Diggelen, Karel van den Bosch, and Mark Anto- nius Neerincx. 2018. Contrastive Explanations for Reinforcement Learning in terms of Expected Consequences. ArXiv abs/1807.08706 (2018). https: //api.semanticscholar.org/CorpusID:49907182

    2018 arXiv

  29. [38]

    Hado van Hasselt, Arthur Guez, and David Silver. 2015. Deep Reinforcement Learning with Double Q-learning. CoRR abs/1509.06461 (2015). arXiv:1509.06461 http://arxiv.org/abs/1509.06461

    2015 arXiv

  30. [39]

    Brijen Thananjeyan, Ashwin Balakrishna, Ugo Rosolia, Felix Li, Rowan McAllister, Joseph E Gonzalez, Sergey Levine, Francesco Borrelli, and Ken Goldberg. 2020. Safety augmented value estimation from demonstrations (saved): Safe deep model- based rl for sparse cost robotic tasks...

    2020

  31. [40]

    Tom Zahavy, Nir Ben-Zrihem, and Shie Mannor. 2016. Graying the black box: Understanding DQNs. In Proceedings of The 33rd International Conference on Machine Learning (Proceedings of Machine Learning Research, Vol. 48) , Maria Flo- rina Balcan and Kilian Q. Weinberger (Eds.). P...

    2016

  32. [44]

    Viégas, and Martin Wattenberg

    Kanit Wongsuphasawat, Daniel Smilkov, James Wexler, Jimbo Wilson, Dandelion Mané, Doug Fritz, Dilip Krishnan, Fernanda B. Viégas, and Martin Wattenberg

  33. [45]

    IEEE Transactions on Visualization and Computer Graphics 24, 1 (2018), 1–12

    Visualizing Dataflow Graphs of Deep Learning Models in TensorFlow. IEEE Transactions on Visualization and Computer Graphics 24, 1 (2018), 1–12. https://doi.org/10.1109/TVCG.2017.2744878

    2018

  34. [2017]

    IEEE Transactions on Visualization and Computer Graphics 23, 1 (2017), 341–350

    Vega-Lite: A Grammar of Interactive Graphics. IEEE Transactions on Visualization and Computer Graphics 23, 1 (2017), 341–350. https://doi.org/10. 1109/TVCG.2016.2599030

    2017

  35. [2018]

    The International journal of robotics research 37, 4-5 (2018), 421–436

    Learning hand-eye coordination for robotic grasping with deep learning and large-scale data collection. The International journal of robotics research 37, 4-5 (2018), 421–436

    2018

  36. [2019]

    https: //api.semanticscholar.org/CorpusID:204898543

    Explainable Reinforcement Learning via Reward Decomposition. https: //api.semanticscholar.org/CorpusID:204898543

  37. [2020]

    https://api.semanticscholar.org/CorpusID: 228898185

    Understanding RL vision. https://api.semanticscholar.org/CorpusID: 228898185

Pith tools

Reviewed August 11, 2026 · model on record in the stance chip above.