Pith. sign in

REVIEW 4 major objections 5 minor 24 references

Out-of-Distribution Detection for Neurosymbolic Autonomous Cyber Agents

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 A transition-probability monitor can flag out-of-distribution states and adversarial strategy switches in discrete-state discrete-action RL cyber agents at runtime.

desk verdict Plausible system demo for discrete-state OOD detection in cyber defense, but the safety-assurance claim is unsupported because the chosen threshold makes the detector a pure novelty detector. read the letter →

arxiv 2412.02875 v1 pith:4E5RJBTQ submitted 2024-12-03 cs.LG cs.AIcs.CR

Ankita Samaddar , Nicholas Potteiger , Xenofon Koutsoukos This is my paper · ORCID
classification cs.LGcs.AIcs.CR
keywords out-of-distributiondetectionprobabilisticneuralnetworkreinforcementlearningcyberdefenseneurosymbolicAIbehaviortreesruntimesafetyassurancediscretestate-actionspaces
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 argues that a reinforcement-learned cyber-defense agent can be made safer at runtime by checking whether each state transition it takes was observed often enough during training. The proposed monitor trains a Probabilistic Neural Network (PNN) on recorded transitions $(s_{t-1},a_{t-1}) \to s_t$, then flags any current state whose estimated transition probability falls below a threshold $\rho$ as out-of-distribution. The claim is that this single mechanism detects anomalous situations and adversarial strategy switches, and that coupling it with a restore action inside the agent's behavior tree turns detection into a concrete safety response. A sympathetic reader would care because the monitor is policy-agnostic for agents with discrete states and discrete actions, and it does not require retraining the defender.

What carries the argument

The load-bearing object is a Probabilistic Neural Network used as a nonparametric empirical transition model. For a query $(s_{t-1},a_{t-1})$, the pattern layer activates only training examples with that same previous state and action, and the summation layer outputs a probability distribution over the distinct successor states seen in training; the usual fourth decision layer is removed. The companion decision rule is the Transition Probability Threshold $\rho$ of Definition 1: a transition is in-distribution only if the successor appears among the PNN's predictions with probability greater than $\rho$. Together they convert memory of training rollouts into a per-timestep safe/OOD verdict.

What would settle it

Run the monitor with $\rho = 0$ on a fresh held-out set of episodes from the same environment and policy used to generate the training data, and count flagged transitions: a non-negligible flag rate would show the detector is responding to finite-sample rarity rather than true distribution shift. Alternatively, inject a harmful transition that was present in the training data; the monitor will label it in-distribution by construction.

Watch

Extended reading notes

Core claim

The central claim is that empirical transition frequency is a sufficient signal for runtime out-of-distribution detection in discrete RL agents. Definition 1 formalizes this: a transition $(s_{t-1},a_{t-1}) \to s_t$ is out-of-distribution under policy $\pi$ when its training-data probability is less than the threshold $\rho$. The PNN learns that distribution by storing every training transition and activating only pattern units whose previous state and action match the current input, so the output layer gives probabilities over observed successor states. In the CybORG CAGE Challenge Scenario 2 evaluation, the monitor with $\rho = 0$ flags 1.5% of episodes against the Meander adversary and 0.1% against Bline, detects a Meander-to-Bline strategy switch at the next timestep, and the added GetSafeAction restore node cuts OOD transitions per episode to between 0 and 2. The paper concludes that the algorithm detects OOD situations under all settings and hence assures safety.

Load-bearing premise

The monitor assumes that any state transition that never (or rarely) appeared in the finite training dataset is genuinely out-of-distribution and unsafe, and that any transition that did appear is safe.

Editorial extensions

If this is right

  • Any discrete-state, discrete-action RL policy can be wrapped with this monitor without modifying the policy itself, since the PNN is trained on observed transitions rather than on policy parameters.
  • An adversarial strategy switch is detectable within one timestep because the new strategy produces transitions that were absent or rare in the training distribution of the old policy.
  • Executing a restore action upon OOD detection lets the agent return to a previously safe state and switch to the control policy trained for the new adversary.
  • With the selected threshold $\rho=0$, the monitor is highly conservative: it flags only transitions that never appeared in 10,000 training episodes, yielding 15 OOD episodes out of 1000 for Meander and 1 out of 1000 for Bline.
  • The PNN remains tractable in this setting because only a small subset of the 52-bit state space is reachable, keeping the output layer small.

Reading between the lines

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

  • The paper equates 'absent from training data' with 'unsafe', which is a coverage assumption rather than a demonstrated fact; a safe-but-rare transition would be flagged OOD, while a harmful transition that happened to appear in training would pass as in-distribution. Pairing the monitor with independent safety labels or a coverage analysis would test this directly.
  • The threshold selection is fragile: Table I shows that $\rho = 10^{-5}$ flags essentially all episodes, so the practical operating point is exactly zero-probability transitions. Recomputing the table with more training data would reveal whether the 0.1-1.5% OOD rate is stable or an artifact of finite sampling.
  • The same transition-frequency idea could be extended toward continuous-state agents by discretizing states or using a density estimator, although the paper explicitly restricts its claim to discrete states and discrete actions.
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

4 major / 5 minor

Summary. The paper proposes an out-of-distribution (OOD) monitoring algorithm for reinforcement-learning-based agents with discrete states and discrete actions. The algorithm learns transition probabilities from training data using a probabilistic neural network (PNN), and at runtime flags a transition (s_{t-1}, a_{t-1}) -> s_t as OOD if its estimated training probability is below a threshold rho. The authors integrate this monitor into an evolving behavior tree (EBT) cyber-defense agent and evaluate it in the CybORG CAGE Challenge Scenario 2 simulator under Meander, Blines, and RedSwitch adversarial strategies. The main claims are that the algorithm detects OOD situations under these settings, that the added GetSafeAction! behavior restores safety after a strategy switch, and that the approach 'can assure safety' in general.

Significance. The problem is well motivated: autonomous cyber-defense agents trained in simulation need runtime monitors that detect situations outside their training distribution and hand them to human experts. The proposed method is simple and, for the discrete-state setting, differentiates this work from continuous-state OOD detectors for RL. The strategy-switch experiments, especially the comparison with and without GetSafeAction!, provide a concrete and suggestive demonstration that a monitor can trigger a recovery action after an adversary changes behavior. However, the paper's central safety claim is not supported by the evidence presented. There is no ground-truth OOD labeling, no false-positive or false-negative analysis, no coverage or generalization analysis, and no baseline comparison. The choice rho=0 turns the detector into a pure novelty detector, which may be a reasonable starting point for detecting distribution shift but is not by itself a safety guarantee. The manuscript also ships a complete algorithm description and experimental setup, which is helpful for reproducibility, although no code is provided.

major comments (4)
  1. [Section VII-B, Table I] The threshold rho is selected from the test data after observing that rho=1e-5 flags 1000 out of 1000 episodes for Meander and 782 out of 1000 for Blines; the paper then sets rho=0 for all subsequent experiments. This is a post hoc choice on the evaluation set, with no held-out validation, no theory, and no sensitivity analysis. Because rho=0 makes Algorithm 1's condition equivalent to 'the transition count in D_train is zero', any transition observed even once is considered in-distribution. This directly contradicts the motivation in Section V-B that confidence requires a transition to occur 'a significantly large number of times.' The authors should either justify rho=0 with a separate validation procedure, report precision/recall over a range of rho on held-out data, or both.
  2. [Section VIII and Problem Statement in Section V-A] The conclusion states that the proposed algorithm 'effectively detects OOD situations under all settings and hence can assure safety,' but the paper provides no evidence that low empirical transition probability is equivalent to unsafety. Two concrete failure modes are not addressed: (1) false negatives, where a dangerous transition that occurred in training has positive empirical probability and will never be flagged; and (2) false positives, where a benign rare transition absent from the finite training set is flagged and triggers GetSafeAction!, possibly disrupting operations. The problem statement asks 'Can we assure safety if the system transitions to any state s' such that Pr((s,a)->s') < rho?', but no answer in the form of a guarantee, bound, or even an empirical safety metric is given. The authors should either provide a rigorous safety argument with explicit assumptions or replace the safety-assurance claim with a more limited claim about detecting transitions unseen in training.
  3. [Section VII-B, Definition 2] The evaluation of OOD episodes is circular: Definition 2 defines an episode as OOD if it contains a transition with Pr < rho, where Pr is the same estimated transition probability used by Algorithm 1. Thus the reported detection counts (15/1000 for Meander and 1/1000 for Blines at rho=0) are largely a restatement of the threshold definition rather than an independent measure of detection quality. The strategy-switch experiments provide a more independent grounding because the RedSwitch scenario uses a held-out red strategy, but these results are reported only qualitatively for five episodes and through aggregate counts in Figures 8-10. The authors should quantify detection delay, false-alarm rate, and accuracy against a labeled set of strategy-switch episodes, and should report results over multiple random seeds with error bars.
  4. [Section VII, Experimental Evaluation] The experimental methodology lacks several standard elements needed to support the paper's claims. No baseline detector is compared (for example, a simple empirical frequency table, a likelihood-based method, or an existing OOD detector for RL); no variance or confidence intervals are reported despite the 1000-episode evaluation; and no separate train/test split for the threshold choice is described. Given the central role of the safety claim, the absence of false-positive and false-negative rates, and the absence of any measure of coverage of the state space by D_train, make it impossible to distinguish the proposed method's behavior from a naive count-based novelty detector. Adding baselines, multiple seeds, and precision/recall or ROC-style evaluation would substantially strengthen the paper.
minor comments (5)
  1. [Section V-B, Algorithm 1 and Figure 3] The PNN input layer is described as having size 1, but the input is the pair (s_{t-1}, a_{t-1}); the figure and text should clarify how the state-action pair is encoded and matched against the pattern layer.
  2. [Throughout] The red agent names 'Meander' and 'Blines' are typeset with an extra space in several places (e.g., 'M eander' and 'B line'); this should be corrected for readability.
  3. [Section V-B, paragraph after Definition 1] The text states that a transition is considered in-distribution if the associated probability is 'greater than' rho, while Algorithm 1 uses '>' in the condition; if rho=0, an observed transition with probability exactly 0 is OOD, which is consistent, but the strictness should be stated consistently and the case of probability exactly rho should be addressed explicitly.
  4. [References] The in-text citation for the Haider et al. work on probabilistic dynamics models appears as [20] in Section V-B, but reference [19] in the bibliography appears to be the Haider et al. AAMAS 2023 paper; the citation numbering should be checked and corrected.
  5. [Figures 6 and 7] The captions refer to median values 'marked in red' in the reward-distribution plots; please ensure the figures actually display these marks clearly and describe the box-plot or violin-plot semantics in the captions.

Circularity Check

2 steps flagged · score 5.0 of 10

The OOD episode counts and the safety conclusion are largely restatements of the paper's own Definition 1 threshold; the strategy-switch scenario provides partial independent content, so circularity is partial rather than total.

  1. self definitional [Section V-A, Definition 1; Section V-B, Algorithm 1]
    "Given a neurosymbolic cyber-agent trained with a policy π, a state transition at timestep t − 1, denoted by (st−1, at−1) → st, is considered to be an out-of-distribution (OOD) transition based on policy π if the probability of occurrence of this transition in the training data is less than threshold ρ, i.e., Pr ((st−1, at−1) → st) < ρ."

    Algorithm 1's decision rule is the same predicate as Definition 1: a transition is in-distribution only if its estimated empirical probability under Dtrain is greater than ρ, and otherwise called OOD. The PNN supplies that empirical probability from observed transition counts. Thus 'the algorithm detects OOD' and 'the transition satisfies Definition 1' are the same operation by construction. Presenting this as an algorithmic result adds no independent evidence that such transitions are anomalous or unsafe; it restates the definition as a runtime check.

  2. self definitional [Section VII-B, Definition 2 and Table I]
    "We define an OOD episode as follows. Definition 2. An episode e is considered to be OOD if there exists at-least one transition in e whose transition probability is less than the Transition Probability Threshold, ρ, i.e., e is an OOD episode, if ∃((st−1, at−1) → st) | Pr((st−1, at−1) → st) < ρ."

    Definition 2 labels an episode OOD using the same threshold predicate and the same PNN estimates that Algorithm 1 uses for detection. Table I then counts episodes satisfying that predicate. No independent OOD labels, ground-truth safety outcomes, precision, recall, or coverage analysis are provided, so the table is a self-consistency count rather than an empirical validation. The subsequent choice of ρ = 0 is made from these self-defined counts, and the later 'effective detection' evidence is therefore partly a restatement of the threshold definition.

full rationale

The paper's central OOD notion is stipulative: Definition 1 defines OOD as low empirical transition probability under the training data, and Algorithm 1 implements exactly that predicate. Consequently, the OOD-episode counts in Table I are not independent measurements of detection quality; they are the definition applied to collected episodes. This is the main circular element, and it also weakens the Section VIII safety claim, which is inferred from those self-defined counts. There is, however, partial independent content. The PNN is fit on 10,000 training episodes, and the known/unknown strategy-switch experiments use a genuinely held-out red-agent behavior (transitions not encountered under the training strategy). The fact that such a switch produces OOD flags is a real distribution-shift response, not purely a tautology. The paper's self-citations to prior EBT work [5] provide the agent, RedSwitch scenario, and simulator integration, but those citations are not load-bearing for the OOD definition itself, and there is no imported uniqueness theorem or machine-checked result that forces the conclusion. On balance, the evaluation's primary metric reduces by construction to the paper's own definition, giving a partial circularity score of 5. The absent external ground truth for OOD/safety is a correctness and validity concern, but the most concrete circularity is the identity among Definition 1, Algorithm 1, and Definition 2's OOD-episode predicate.

Assumptions & free parameters 1 free parameters · 5 assumptions · 0 invented entities

The central claim rests on five domain assumptions: the POMDP model, completeness of the training transition distribution, the semantic link between low transition probability and unsafety, PNN accuracy, and simulator validity. The only numeric free parameter is the threshold rho, and it is selected from the test data. No new physical or conceptual entities are introduced beyond the formal definitions of OOD transitions and the threshold.

free parameters (1)
  • Transition Probability Threshold rho = 0
    Chosen from Table I: rho=1e-5 and above flag all 1000 episodes as OOD against Meander, so rho is set to 0 for all subsequent experiments. This is a data-driven operating point, not a principled threshold.
assumptions (5)
  • domain assumption The cyber environment is accurately modeled as a discrete-time POMDP with stationary transition probabilities under a fixed red strategy.
    Section V-A formulates the system as M=(S,A,T,R,mu0) and assumes the transition kernel is fixed during training and monitoring.
  • domain assumption The empirical transition distribution from 10,000 episodes is complete enough that any unobserved transition is genuinely out-of-distribution.
    Definition 1 and the choice rho=0 rely on this coverage assumption; no statistical guarantee or support analysis is provided.
  • domain assumption Low transition probability under the training distribution is a reliable signal of unsafe situations and of adversarial strategy changes.
    This links the threshold test in Section V-B to the safety assurance claim in Section VIII, but the paper provides no independent ground truth for OOD labels.
  • domain assumption The PNN estimates transition probabilities accurately enough for safety decisions without calibration or uncertainty quantification.
    The training phase in Section V-B reports no calibration, confidence intervals, or error analysis for the PNN estimates.
  • domain assumption Simulation results in CybORG CAGE Challenge Scenario 2 are representative of real network defense behavior.
    All experiments use the simulator, and Section VIII acknowledges that real testbeds may behave differently.

how reviews work

0 comments
Cite this review

Pith. "Pith review of Out-of-Distribution Detection for Neurosymbolic Autonomous Cyber Agents." pith.science (2026). https://pith.science/paper/4E5RJBTQ

@misc{pith2026241202875,
  author       = {Pith},
  title        = {Pith review of: Out-of-Distribution Detection for Neurosymbolic Autonomous Cyber Agents},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/4E5RJBTQ}},
  note         = {Machine review of arXiv:2412.02875}
}
read the original abstract

Autonomous agents for cyber applications take advantage of modern defense techniques by adopting intelligent agents with conventional and learning-enabled components. These intelligent agents are trained via reinforcement learning (RL) algorithms, and can learn, adapt to, reason about and deploy security rules to defend networked computer systems while maintaining critical operational workflows. However, the knowledge available during training about the state of the operational network and its environment may be limited. The agents should be trustworthy so that they can reliably detect situations they cannot handle, and hand them over to cyber experts. In this work, we develop an out-of-distribution (OOD) Monitoring algorithm that uses a Probabilistic Neural Network (PNN) to detect anomalous or OOD situations of RL-based agents with discrete states and discrete actions. To demonstrate the effectiveness of the proposed approach, we integrate the OOD monitoring algorithm with a neurosymbolic autonomous cyber agent that uses behavior trees with learning-enabled components. We evaluate the proposed approach in a simulated cyber environment under different adversarial strategies. Experimental results over a large number of episodes illustrate the overall efficiency of our proposed approach.

Figures

Figures reproduced from arXiv: 2412.02875 by the authors.

Figure 1
Figure 1. Network architecture of CybORG Cage Challenge Scenario 2 with three subnets [7]; Subnet 1 with five hosts, Subnet 2 with three enterprise servers [PITH_FULL_IMAGE:figures/full_fig_p003_1.png] view at source ↗
Figure 2
Figure 2. Optimal Behavior Tree for Autonomous Cyber-Defense [5] [PITH_FULL_IMAGE:figures/full_fig_p004_2.png] view at source ↗
Figure 3
Figure 3. A PNN with input layer of size 1, pattern layer of size [PITH_FULL_IMAGE:figures/full_fig_p005_3.png] view at source ↗
Figures from the paper (7 more)
Figure 4
Figure 4. Figure 4: Updated Behavior Tree for OOD Monitoring; The newly added nodes and connections are marked in red in the updated BT. [PITH_FULL_IMAGE:figures/full_fig_p006_4.png]
Figure 5
Figure 5. Figure 5: Software Architecture for OOD Monitoring in autonomous cyber [PITH_FULL_IMAGE:figures/full_fig_p006_5.png]
Figure 6
Figure 6. Figure 6: Reward Distribution under different Transition Probability Thresholds [PITH_FULL_IMAGE:figures/full_fig_p007_6.png]
Figure 7
Figure 7. Figure 7: Reward Distribution under different Transition Probability Thresholds [PITH_FULL_IMAGE:figures/full_fig_p007_7.png]
Figure 8
Figure 8. Figure 8: Five episodes under strategy switching ( [PITH_FULL_IMAGE:figures/full_fig_p008_8.png]
Figure 9
Figure 9. Figure 9: Number of OOD transitions per episode under strategy switching [PITH_FULL_IMAGE:figures/full_fig_p008_9.png]
Figure 10
Figure 10. Figure 10: Number of OOD transitions per episode under known strategy [PITH_FULL_IMAGE:figures/full_fig_p009_10.png]

Discussion (0). Continue with ORCID to comment.

Reference graph

Works this paper leans on

24 extracted references · 21 canonical work pages

  1. [1]

    Task-driven out-of-distribution detection with statistical guarantees for robot learning,

    A. Farid, S. Veer, and A. Majumdar, “Task-driven out-of-distribution detection with statistical guarantees for robot learning,” in 5th Annual Conference on Robot Learning , 2021

    work page 2021

  2. [2]

    Efficient out-of-distribution detection using latent space of β-vae for cyber-physical systems,

    S. Ramakrishna, Z. Rahiminasab, G. Karsai, A. Easwaran, and A. Dubey, “Efficient out-of-distribution detection using latent space of β-vae for cyber-physical systems,” ACM Transactions on Cyber-Physical Systems, vol. 6, Apr. 2022

    work page 2022

  3. [3]

    Real-time out-of-distribution detection in learning-enabled cyber-physical systems,

    F. Cai and X. Koutsoukos, “Real-time out-of-distribution detection in learning-enabled cyber-physical systems,” in 2020 ACM/IEEE 11th International Conference on Cyber-Physical Systems (ICCPS) , pp. 174– 183, 2020

    work page 2020

  4. [4]

    Colledanchise and P

    M. Colledanchise and P. ¨Ogren, Behavior Trees in Robotics and AI . CRC Press, jul 2018

    work page 2018

  5. [5]

    Design- ing robust cyber-defense agents with evolving behavior trees,

    N. Potteiger, A. Samaddar, H. Bergstrom, and X. Koutsoukos, “Design- ing robust cyber-defense agents with evolving behavior trees,” in arxiv, 2024

    work page 2024

  6. [6]

    Anomaly detection in thermal power plant using probabilistic neural network,

    A. Hajdarevic, I. Dzananovic, L. Banjanovic-Mehmedovic, and F. Mehmedovic, “Anomaly detection in thermal power plant using probabilistic neural network,” in 2015 38th International Convention on Information and Communication Technology, Electronics and Micro- electronics (MIPRO), pp. 1118–1123, 2015

    work page 2015

  7. [7]

    Cyber autonomy gym for experimentation challenge 2

    “Cyber autonomy gym for experimentation challenge 2.” https://github. com/cage-challenge/cage-challenge-2, 2022. Created by Maxwell Standen, David Bowman, Son Hoang, Toby Richer, Martin Lucas, Richard Van Tassel, Phillip Vu, Mitchell Kiely

    work page 2022

  8. [8]

    On autonomous agents in a cyber defence environment,

    M. Kiely, D. Bowman, M. Standen, and C. Moir, “On autonomous agents in a cyber defence environment,” ArXiv, vol. abs/2309.07388, 2023

    arXiv 2023

Show all 24 references
  1. [9]

    Autonomous net- work defence using reinforcement learning,

    M. Foley, C. Hicks, K. Highnam, and V . Mavroudis, “Autonomous net- work defence using reinforcement learning,” in Proceedings of the 2022 ACM on Asia Conference on Computer and Communications Security , ASIA CCS ’22, (New York, NY , USA), p. 1252–1254, Association for Computi...

    2022

  2. [10]

    Beyond cage: Investigating generalization of learned autonomous network defense policies,

    M. Wolk, A. Applebaum, C. Dennler, P. Dwyer, M. Moskowitz, H. Nguyen, N. Nichols, N. Park, P. Rachwalski, F. Rau, and A. Web- ster, “Beyond cage: Investigating generalization of learned autonomous network defense policies,” ArXiv, vol. abs/2211.15557, 2022

    2022 arXiv

  3. [11]

    Net- work environment design for autonomous cyberdefense,

    A. Molina-Markham, C. Miniter, B. Powell, and A. Ridley, “Net- work environment design for autonomous cyberdefense,” ArXiv, vol. abs/2103.07583, 2021

    2021 arXiv

  4. [12]

    Neurosymbolic ai in cybersecurity: Bridging pattern recognition and symbolic reasoning,

    B. Jalaian and N. D. Bastian, “Neurosymbolic ai in cybersecurity: Bridging pattern recognition and symbolic reasoning,” in MILCOM 2023 - 2023 IEEE Military Communications Conference (MILCOM), pp. 268– 273, 2023

    2023

  5. [13]

    Evaluating behaviour tree integration in the option critic framework in starcraft 2 mini-games with training restricted by consumer level hardware,

    F. Lundberg, “Evaluating behaviour tree integration in the option critic framework in starcraft 2 mini-games with training restricted by consumer level hardware,” Master’s thesis, KTH, School of Electrical Engineering and Computer Science (EECS), 2022

    2022

  6. [14]

    Mixed deep reinforcement learning- behavior tree for intelligent agents design,

    L. Li, L. Wang, Y . Li, and J. Sheng, “Mixed deep reinforcement learning- behavior tree for intelligent agents design,” in International Conference on Agents and Artificial Intelligence , 2021

    2021

  7. [15]

    The option-critic architecture,

    P.-L. Bacon, J. Harb, and D. Precup, “The option-critic architecture,” in Proceedings of the Thirty-First AAAI Conference on Artificial Intel- ligence, AAAI’17, p. 1726–1734, AAAI Press, 2017

    2017

  8. [16]

    Can autonomous vehicles identify, recover from, and adapt to distribution shifts?,

    A. Filos, P. Tigas, R. T. McAllister, N. Rhinehart, S. Levine, and Y . Gal, “Can autonomous vehicles identify, recover from, and adapt to distribution shifts?,” in International Conference on Machine Learning , 2020

    2020

  9. [17]

    Unified out-of-distribution detection: A model-specific perspective,

    A. Reza and C. Wei-Lun, “Unified out-of-distribution detection: A model-specific perspective,” 2023 IEEE/CVF International Conference on Computer Vision (ICCV) , pp. 1453–1463, 2023

    2023

  10. [18]

    Full-spectrum out-of-distribution detec- tion,

    J. Yang, K. Zhou, and Z. Liu, “Full-spectrum out-of-distribution detec- tion,” International Journal of Computer Vision, vol. 131, p. 2607–2622, June 2023

    2023

  11. [19]

    Out-of-distribution detection for reinforcement learning agents with probabilistic dynamics models,

    T. Haider, K. Roscher, F. Schmoeller da Roza, and S. G ¨unnemann, “Out-of-distribution detection for reinforcement learning agents with probabilistic dynamics models,” inProceedings of the 2023 International Conference on Autonomous Agents and Multiagent Systems , AAMAS ’23, p...

    2023

  12. [20]

    Rethinking out-of-distribution detection for reinforcement learning: Advancing methods for evaluation and detection,

    L. Nasvytis, K. Sandbrink, J. Foerster, T. Franzmeyer, and C. S. de Witt, “Rethinking out-of-distribution detection for reinforcement learning: Advancing methods for evaluation and detection,” 2024

    2024

  13. [21]

    Pas: Probably approximate safety verifi- cation of reinforcement learning policy using scenario optimization,

    A. J. Singh and A. Easwaran, “Pas: Probably approximate safety verifi- cation of reinforcement learning policy using scenario optimization,” in Proceedings of the 23rd International Conference on Autonomous Agents and Multiagent Systems , AAMAS ’24, p. 1745–1753, International...

    2024

  14. [22]

    CybORG: A Gym for the Development of Autonomous Cyber Agents , 2021

    2021

  15. [23]

    M. L. Puterman, Markov Decision Processes: Discrete Stochastic Dy- namic Programming. John Wiley & Sons, 2014

    2014

  16. [24]

    Pytrees

    S. Reality, “Pytrees.” https://github.com/splintered-reality/py trees, 2023

    2023

Pith tools

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