pith. sign in

arxiv: 2506.05171 · v3 · submitted 2025-06-05 · 📡 eess.SY · cs.AI· cs.SY

Towards provable probabilistic safety for scalable embodied AI systems

Linxuan He , Lingxiang Fan , Qing-Shan Jia , Ang Li , Hongyan Sang , Ling Wang , Guanghui Wen , Jiwen Lu
show 8 more authors
Tao Zhang Jie Zhou Yi Zhang Yisen Wang Peng Wei Zhongyuan Wang Henry X. Liu Shuo Feng
This is my paper

Pith reviewed 2026-05-19 10:57 UTC · model grok-4.3

classification 📡 eess.SY cs.AIcs.SY
keywords provable probabilistic safetyembodied AIsafety verificationscalable deploymentautonomous systemsstatistical methodssafety-critical applicationsparadigm shift
0
0 comments X

The pith

Embodied AI systems achieve scalable safety by shifting to provable probabilistic guarantees instead of deterministic verification.

A machine-rendered reading of the paper's core claim, the machinery that carries it, and where it could break.

Embodied AI systems like autonomous vehicles and robots struggle with safety because failures are rare yet corner cases are too complex for exhaustive deterministic checks across all scenarios. The paper proposes replacing that ideal with provable probabilistic safety, which supplies verifiable guarantees while allowing gradual progress toward a defined probabilistic safety boundary on overall performance. This approach draws on statistical methods to improve practicality and scalability where pure verification fails. A reader would care because it directly addresses the barrier to large-scale deployment in safety-critical domains by offering a workable middle path between theory and real-world use.

Core claim

The paper claims that a paradigm shift to provable probabilistic safety, integrating provable guarantees with progressive achievement toward a probabilistic safety boundary on overall system performance, makes safety assurance feasible and enables embodied AI systems to be deployed at scale. This paradigm leverages statistical methods for better feasibility while a well-defined boundary supports practical large-scale adoption in complex environments.

What carries the argument

Provable probabilistic safety, the mechanism that combines provable guarantees with progressive movement toward a probabilistic safety boundary on system performance.

If this is right

  • Embodied AI systems become deployable at scale once a probabilistic safety boundary is established and met.
  • Statistical methods gain a central role in making safety verification practical for large systems.
  • A roadmap of challenges and solutions guides implementation of the new safety approach.
  • Theoretical safety assurance connects more directly to practical deployment in domains such as vehicles and robotics.

Where Pith is reading between the lines

These are editorial extensions of the paper, not claims the author makes directly.

  • Safety certification processes for physical AI systems could shift emphasis toward measurable probabilistic bounds.
  • Testing protocols might prioritize sampling strategies that bound rare events rather than exhaustive scenario coverage.
  • The approach could generalize to other domains with rare but high-impact failures where full determinism is intractable.

Load-bearing premise

Statistical and probabilistic methods can deliver meaningful provable guarantees for complex embodied systems despite the rarity and complexity of corner cases.

What would settle it

A deployed embodied AI system that satisfies the probabilistic safety boundary yet exhibits a higher-than-expected real-world failure rate in operation would falsify the central claim.

read the original abstract

Embodied AI systems, comprising AI models and physical plants, are increasingly prevalent across various applications. Due to the rarity of system failures, ensuring their safety in complex operating environments remains a major challenge, which severely hinders their large-scale deployment in safety-critical domains, such as autonomous vehicles, medical devices, and robotics. While achieving provable deterministic safety-verifying system safety across all possible scenarios-remains theoretically ideal, the rarity and complexity of corner cases make this approach impractical for scalable embodied AI systems. Instead, empirical safety evaluation is employed as an alternative, but the absence of provable guarantees imposes significant limitations. To address these issues, we argue for a paradigm shift to provable probabilistic safety that integrates provable guarantees with progressive achievement toward a probabilistic safety boundary on overall system performance. The new paradigm better leverages statistical methods to enhance feasibility and scalability, and a well-defined probabilistic safety boundary enables embodied AI systems to be deployed at scale. In this Perspective, we outline a roadmap for provable probabilistic safety, along with corresponding challenges and potential solutions. By bridging the gap between theoretical safety assurance and practical deployment, this Perspective offers a pathway toward safer, large-scale adoption of embodied AI systems in safety-critical applications.

Editorial analysis

A structured set of objections, weighed in public.

Desk editor's note, referee report, simulated authors' rebuttal, and a circularity audit. Tearing a paper down is the easy half of reading it; the pith above is the substance, this is the friction.

Referee Report

2 major / 2 minor

Summary. The paper argues that deterministic safety verification is impractical for embodied AI systems due to rare corner cases, while purely empirical methods lack provable guarantees; it proposes a paradigm shift to 'provable probabilistic safety' that integrates provable guarantees with progressive achievement of a probabilistic safety boundary on overall system performance, outlines a high-level roadmap with challenges and solutions, and claims this enables scalable deployment in domains such as autonomous vehicles and robotics.

Significance. If concrete mechanisms for achieving the proposed integration of provable guarantees and probabilistic boundaries can be developed, the perspective could help address a genuine scalability barrier in safety-critical embodied AI. The manuscript correctly identifies limitations of existing deterministic and empirical approaches and gives credit to the need for statistical methods, but remains conceptual without new technical results, derivations, or examples.

major comments (2)
  1. [Abstract] Abstract: The central claim that 'provable probabilistic safety' integrates provable guarantees with a probabilistic safety boundary 'better leverages statistical methods to enhance feasibility and scalability' is presented as a paradigm shift without any specific mechanism, algorithm, bound, or illustrative example showing how such integration would deliver meaningful provable guarantees for complex embodied systems; this is load-bearing for the feasibility argument.
  2. [Roadmap] Roadmap section: The outline of challenges and potential solutions for the probabilistic safety boundary does not specify any verification procedure, statistical test, or progressive achievement criterion that would allow falsification or validation of the claimed guarantees, leaving the roadmap at too high a level to support the scalability conclusion.
minor comments (2)
  1. [Introduction] The term 'provable probabilistic safety' is introduced without a formal definition or distinction from existing probabilistic verification frameworks; adding a short clarifying paragraph would improve precision.
  2. Several references to 'statistical methods' and 'empirical safety evaluation' would benefit from citing specific prior work in probabilistic model checking or statistical model checking for cyber-physical systems.

Simulated Author's Rebuttal

2 responses · 0 unresolved

We thank the referee for their constructive comments on our Perspective paper. We appreciate the acknowledgment that the work correctly identifies limitations of existing approaches. As a Perspective article, the manuscript is intentionally conceptual and high-level, proposing a paradigm and roadmap rather than new technical results. Below we respond point by point to the major comments and indicate planned revisions.

read point-by-point responses

  1. Referee: [Abstract] Abstract: The central claim that 'provable probabilistic safety' integrates provable guarantees with a probabilistic safety boundary 'better leverages statistical methods to enhance feasibility and scalability' is presented as a paradigm shift without any specific mechanism, algorithm, bound, or illustrative example showing how such integration would deliver meaningful provable guarantees for complex embodied systems; this is load-bearing for the feasibility argument.

    Authors: We agree that the manuscript presents the integration at a conceptual level without specific mechanisms or examples, which is consistent with its nature as a Perspective outlining a new paradigm rather than deriving technical results. The claim is grounded in the observation that combining formal guarantees with statistical methods can address rare corner cases more scalably than exhaustive deterministic verification. To address the concern about the load-bearing nature of the claim, we will revise the abstract to explicitly note that the paper offers a high-level framework and that concrete algorithms and bounds are directions for future research. This revision will clarify expectations while preserving the core argument. revision: yes

  2. Referee: [Roadmap] Roadmap section: The outline of challenges and potential solutions for the probabilistic safety boundary does not specify any verification procedure, statistical test, or progressive achievement criterion that would allow falsification or validation of the claimed guarantees, leaving the roadmap at too high a level to support the scalability conclusion.

    Authors: The referee is correct that the roadmap section remains high-level and does not detail specific verification procedures or statistical tests. This level of abstraction is appropriate for a Perspective paper whose purpose is to identify challenges and potential solution directions rather than to validate a complete methodology. To better support the scalability discussion, we will make a partial revision by adding references to relevant statistical validation techniques (such as sequential testing or PAC-style bounds) from the literature and by elaborating on how progressive achievement of a safety boundary might be assessed empirically while retaining formal elements. We will not claim that these additions constitute new results. revision: partial

Circularity Check

0 steps flagged

No significant circularity; perspective paper with no derivations

full rationale

The manuscript is a perspective article that proposes a high-level conceptual shift toward 'provable probabilistic safety' and outlines a roadmap of challenges and solutions. It contains no mathematical derivations, equations, fitted parameters, or technical proofs. The central argument rests on the acknowledged impracticality of deterministic verification for rare corner cases and the feasibility of integrating statistical methods with probabilistic boundaries, without any load-bearing step that reduces to a self-definition, fitted input, or self-citation chain. Because no derivation chain exists, no circular reduction can be exhibited. The paper is self-contained as a conceptual proposal and does not claim quantitative predictions or uniqueness theorems that would require external verification.

Axiom & Free-Parameter Ledger

0 free parameters · 1 axioms · 0 invented entities

The perspective relies on the domain assumption that probabilistic bounds can be made rigorous enough for safety-critical deployment without providing supporting derivations or evidence.

axioms (1)
  • domain assumption Statistical methods can provide provable guarantees that are both feasible and sufficient for safety-critical embodied AI.
    Invoked in the abstract when contrasting deterministic verification with probabilistic approaches.

pith-pipeline@v0.9.0 · 5795 in / 1030 out tokens · 27264 ms · 2026-05-19T10:57:04.985755+00:00 · methodology

discussion (0)

Sign in with ORCID, Apple, or X to comment. Anyone can read and Pith papers without signing in.

Lean theorems connected to this paper

Citations machine-checked in the Pith Canon. Every link opens the source theorem in the public Lean library.

What do these tags mean?
matches
The paper's claim is directly supported by a theorem in the formal canon.
supports
The theorem supports part of the paper's argument, but the paper may add assumptions or extra steps.
extends
The paper goes beyond the formal theorem; the theorem is a base layer rather than the whole result.
uses
The paper appears to rely on the theorem as machinery.
contradicts
The paper's claim conflicts with a theorem or certificate in the canon.
unclear
Pith found a possible connection, but the passage is too broad, indirect, or ambiguous to say the theorem truly supports the claim.

Reference graph

Works this paper leans on

158 extracted references · 158 canonical work pages · 5 internal anchors

  1. [1]

    Silver, D. et al. Mastering the game of go with deep neural networks and tree search. Nature 529, 484–489 (2016)

    work page 2016

  2. [2]

    Vaswani, A. et al. Attention is all you need. In Advances in Neural Information Processing Systems (NeurIPS) (2017)

    work page 2017

  3. [3]

    Zhao, W. X. et al. A survey of large language models. arXiv preprint arXiv:2303.18223 (2023)

    work page internal anchor Pith review Pith/arXiv arXiv 2023

  4. [4]

    & Fei-Fei, L

    Gupta, A., Savarese, S., Ganguli, S. & Fei-Fei, L. Embodied intelligence via learning and evolution. Nat. Commun. 12, 5721 (2021)

    work page 2021

  5. [5]

    Liu, Y . et al. Aligning cyber space with physical world: A comprehensive survey on embodied AI. arXiv preprint arXiv:2407.06886 (2024)

    work page arXiv 2024

  6. [6]

    Turing, A. M. Computing machinery and intelligence (Springer, 2009)

    work page 2009

  7. [7]

    An overview of catastrophic ai risks

    Hendrycks, D., Mazeika, M. & Woodside, T. An overview of catastrophic AI risks. arXiv preprint arXiv:2306.12001 (2023)

    work page arXiv 2023

  8. [8]

    Zhou, L. et al. Larger and more instructable language models become less reliable. Nature 634, 61–68 (2024)

    work page 2024

  9. [9]

    & Gal, Y

    Farquhar, S., Kossen, J., Kuhn, L. & Gal, Y . Detecting hallucinations in large language models using semantic entropy.Nature 630, 625–630 (2024)

    work page 2024

  10. [10]

    Zhu, Z., Wu, B., Zhang, Z., Han, L. & Wu, B. Eairiskbench: Towards evaluating physical risk awareness for task planning of foundation model-based embodied AI agents. arXiv preprint arXiv:2408.04449 (2024)

    work page arXiv 2024

  11. [11]

    Managing extreme AI risks amid rapid progress

    Bengio, Y .et al. Managing extreme AI risks amid rapid progress. Science 384, 842–845 (2024)

    work page 2024

  12. [12]

    Li, L. et al. Embodied intelligence in mining: Leveraging multi-modal large language model for autonomous driving in mines. IEEE Transactions on Intell. Veh.(2024)

    work page 2024

  13. [13]

    Moor, M. et al. Foundation models for generalist medical artificial intelligence. Nature 616, 259–265 (2023)

    work page 2023

  14. [14]

    Tu, T. et al. Towards conversational diagnostic artificial intelligence. Nature 1–9 (2025)

    work page 2025

  15. [15]

    F., Martin, C

    Nygaard, T. F., Martin, C. P., Torresen, J., Glette, K. & Howard, D. Real-world embodied AI through a morphologically adaptive quadruped robot. Nat. Mach. Intell. 3, 410–419 (2021). 11

    work page 2021

  16. [16]

    Incident number 4: Uber A V killed pedestrian in Arizona

    AI Incident Database, Responsible AI Collaborative. Incident number 4: Uber A V killed pedestrian in Arizona. https: //incidentdatabase.ai/cite/726 (2018)

    work page 2018

  17. [17]

    Incident number 638: Fatal crash involving tesla full self-driving claims employee’s life

    AI Incident Database, Responsible AI Collaborative. Incident number 638: Fatal crash involving tesla full self-driving claims employee’s life. https://incidentdatabase.ai/cite/726 (2022)

    work page 2022

  18. [18]

    Incident 599: Stacking robot fatally crushes employee in South Korea

    AI Incident Database, Responsible AI Collaborative. Incident 599: Stacking robot fatally crushes employee in South Korea. https://incidentdatabase.ai/cite/599 (2023)

    work page 2023

  19. [19]

    Incident 68: Security robot drowns itself in a fountain

    AI Incident Database, Responsible AI Collaborative. Incident 68: Security robot drowns itself in a fountain. https: //incidentdatabase.ai/cite/68 (2017)

    work page 2017

  20. [20]

    Generating robot constitutions & benchmarks for semantic safety,

    Sermanet, P., Majumdar, A., Irpan, A., Kalashnikov, D. & Sindhwani, V . Generating robot constitutions & benchmarks for semantic safety. arXiv preprint arXiv:2503.08663 (2025)

    work page arXiv 2025

  21. [21]

    Dynamic programming

    Bellman, R. Dynamic programming. Science 153, 34–37 (1966)

    work page 1966

  22. [22]

    Liu, H. X. & Feng, S. Curse of rarity for autonomous vehicles. Nat. Commun. 15, 4808 (2024)

    work page 2024

  23. [23]

    Feng, S. et al. Dense reinforcement learning for safety validation of autonomous vehicles. Nature 615, 620–627 (2023)

    work page 2023

  24. [24]

    Bai, R. et al. Accurately predicting probabilities of safety-critical rare events for intelligent systems. In 2024 IEEE 20th International Conference on Automation Science and Engineering (CASE), 3243–3249 (IEEE, 2024)

    work page 2024

  25. [25]

    Mehdipour, N., Althoff, M., Tebbens, R. D. & Belta, C. Formal methods to comply with rules of the road in autonomous driving: State of the art and grand challenges. Automatica 152, 110692 (2023)

    work page 2023

  26. [26]

    A., Dixon, C

    Luckcuck, M., Farrell, M., Dennis, L. A., Dixon, C. & Fisher, M. Formal specification and verification of autonomous robotic systems: A survey. ACM Comput. Surv. (CSUR) 52, 1–41 (2019)

    work page 2019

  27. [27]

    Liu, C. et al. Algorithms for verifying deep neural networks. Foundations Trends Optim.4, 244–404 (2021)

    work page 2021

  28. [28]

    Ames, A. D. et al. Control barrier functions: Theory and applications. In 2019 18th European Control Conference (ECC), 3420–3431 (IEEE, 2019)

    work page 2019

  29. [29]

    & Zhu, Q

    Huang, C., Fan, J., Li, W., Chen, X. & Zhu, Q. Reachnn: Reachability analysis of neural-network controlled systems. ACM Transactions on Embed. Comput. Syst. (TECS) 18, 1–22 (2019)

    work page 2019

  30. [30]

    Zhao, Q. et al. Verifying neural network controlled systems using neural networks. InProceedings of the 25th ACM International Conference on Hybrid Systems: Computation and Control, 1–11 (2022)

    work page 2022

  31. [31]

    Tran, H.-D. et al. Star-based reachability analysis of deep neural networks. In Formal Methods–The Next 30 Years: Third World Congress, FM 2019, Porto, Portugal, October 7–11, 2019, Proceedings 3, 670–686 (Springer, 2019)

    work page 2019

  32. [32]

    & Althoff, M

    Kochdumper, N., Krasowski, H., Wang, X., Bak, S. & Althoff, M. Provably safe reinforcement learning via action projection using reachability analysis and polynomial zonotopes. IEEE Open J. Control. Syst. 2, 79–92 (2023)

    work page 2023

  33. [33]

    Xiao, W. et al. Barriernet: Differentiable control barrier functions for learning of safe robot control. IEEE Transactions on Robotics 39, 2289–2307 (2023)

    work page 2023

  34. [34]

    Dalrymple, D. et al. Towards guaranteed safe AI: A framework for ensuring robust and reliable AI systems. arXiv preprint arXiv:2405.06624 (2024)

    work page arXiv 2024

  35. [35]

    Liu, P., Yang, R. & Xu, Z. How safe is safe enough for self-driving vehicles? Risk Analysis 39, 315–325 (2019)

    work page 2019

  36. [36]

    Certification specifications for normal, utility, aerobatic, and commuter category aeroplanes—cs-23

    European Aviation Safety Agency. Certification specifications for normal, utility, aerobatic, and commuter category aeroplanes—cs-23. Amendment 3, 20–32 (2012)

    work page 2012

  37. [37]

    Anderson, C. W. Learning to control an inverted pendulum using neural networks. IEEE Control. Syst. Mag. 9, 31–37 (1989)

    work page 1989

  38. [38]

    Campbell, M., Hoane Jr, A. J. & Hsu, F.-h. Deep blue. Artif. intelligence 134, 57–83 (2002)

    work page 2002

  39. [39]

    Jumper, J. et al. Highly accurate protein structure prediction with alphafold. Nature 596, 583–589 (2021)

    work page 2021

  40. [40]

    Abramson, J. et al. Accurate structure prediction of biomolecular interactions with alphafold 3. Nature 1–3 (2024)

    work page 2024

  41. [41]

    Mirhoseini, A. et al. A graph placement methodology for fast chip design. Nature 594, 207–212 (2021)

    work page 2021

  42. [42]

    Silver, D. et al. Mastering the game of go without human knowledge. Nature 550, 354–359 (2017)

    work page 2017

  43. [43]

    Burger, B. et al. A mobile robotic chemist. Nature 583, 237–241 (2020)

    work page 2020

  44. [44]

    Dai, T. et al. Autonomous mobile robots for exploratory synthetic chemistry. Nature 1–8 (2024)

    work page 2024

  45. [45]

    Kaufmann, E. et al. Champion-level drone racing using deep reinforcement learning. Nature 620, 982–987 (2023). 12

    work page 2023

  46. [46]

    Mohamad, M. A. & Sapsis, T. P. Sequential sampling strategy for extreme event statistics in nonlinear dynamical systems. Proc. Natl. Acad. Sci. 115, 11138–11143 (2018)

    work page 2018

  47. [47]

    & Zeilinger, M

    Hewing, L. & Zeilinger, M. N. Scenario-based probabilistic reachable sets for recursively feasible stochastic model predictive control. IEEE Control. Syst. Lett. 4, 450–455 (2019)

    work page 2019

  48. [48]

    & Easwaran, A

    Xue, B., Fränzle, M., Zhao, H., Zhan, N. & Easwaran, A. Probably approximate safety verification of hybrid dynamical systems. In International Conference on Formal Engineering Methods, 236–252 (Springer, 2019)

    work page 2019

  49. [49]

    Monte Carlo sampling methods

    Shapiro, A. Monte Carlo sampling methods. Handbooks in operations research and management science 10, 353–425 (2003)

    work page 2003

  50. [50]

    Wang, K. et al. Generative adversarial networks: Introduction and outlook. IEEE/CAA J. Autom. Sinica 4, 588–598 (2017)

    work page 2017

  51. [51]

    Unsolved Problems in ML Safety

    Hendrycks, D., Carlini, N., Schulman, J. & Steinhardt, J. Unsolved problems in ML safety. arXiv preprint arXiv:2109.13916 (2021)

    work page internal anchor Pith review Pith/arXiv arXiv 2021

  52. [52]

    Bensalem, S. et al. What, indeed, is an achievable provable guarantee for learning-enabled safety-critical systems. InInternational Conference on Bridging the Gap between AI and Reality, 55–76 (Springer, 2023)

    work page 2023

  53. [53]

    A., Sadigh, D

    Seshia, S. A., Sadigh, D. & Sastry, S. S. Toward verified artificial intelligence. Commun. ACM 65, 46–55 (2022)

    work page 2022

  54. [54]

    & Omohundro, S

    Tegmark, M. & Omohundro, S. Provably safe systems: The only path to controllable AGI. arXiv preprint arXiv:2309.01933 (2023)

    work page arXiv 2023

  55. [55]

    On the trustworthiness of generative foundation models: Guideline, assessment, and perspective

    Huang, Y .et al. On the trustworthiness of generative foundation models: Guideline, assessment, and perspective. arXiv preprint arXiv:2502.14296 (2025)

    work page internal anchor Pith review arXiv 2025

  56. [56]

    A survey on large language model (LLM) security and privacy: The good, the bad, and the ugly

    Yao, Y .et al. A survey on large language model (LLM) security and privacy: The good, the bad, and the ugly. High-Confidence Comput. 100211 (2024)

    work page 2024

  57. [57]

    Liu, Q. et al. A survey on security threats and defensive techniques of machine learning: A data driven view. IEEE Access 6, 12103–12117 (2018)

    work page 2018

  58. [58]

    & Work, D

    Gunter, G., Nice, M., Bunting, M., Sprinkle, J. & Work, D. Can control barrier functions keep automated vehicles safe in live freeway traffic? In Proceedings of the ACM/IEEE 16th International Conference on Cyber-Physical Systems (with CPS-IoT Week 2025), 1–10 (2025)

    work page 2025

  59. [59]

    Probably approximately correct: Nature’s algorithms for learning and prospering in a complex world(Basic Books, 2013)

    Valiant, L. Probably approximately correct: Nature’s algorithms for learning and prospering in a complex world(Basic Books, 2013)

    work page 2013

  60. [60]

    & Redmill, K

    Weng, B., Capito, L., Ozguner, U. & Redmill, K. Towards guaranteed safety assurance of automated driving systems with scenario sampling: An invariant set perspective. IEEE Transactions on Intell. Veh.7, 638–651 (2021)

    work page 2021

  61. [61]

    & Arcak, M

    Devonport, A. & Arcak, M. Estimating reachable sets with scenario optimization. In Learning for Dynamics and Control, 75–84 (PMLR, 2020)

    work page 2020

  62. [62]

    Larsen, K. G. & Legay, A. Statistical model checking: Past, present, and future. In Leveraging Applications of Formal Methods, Verification and Validation: Foundational Techniques: 7th International Symposium, ISoLA 2016, Imperial, Corfu, Greece, October 10–14, 2016, Proceedings, Part I 7, 3–15 (Springer, 2016)

    work page 2016

  63. [63]

    & Pajic, M

    Wang, Y ., Zarei, M., Bonakdarpour, B. & Pajic, M. Statistical verification of hyperproperties for cyber-physical systems.ACM Transactions on Embed. Comput. Syst. (TECS) 18, 1–23 (2019)

    work page 2019

  64. [64]

    & Haddad, S

    Barbot, B., Bérard, B., Duplouy, Y . & Haddad, S. Statistical model-checking for autonomous vehicle safety validation. In Conference SIA Simulation Numérique (2017)

    work page 2017

  65. [65]

    & Jungers, R

    Wang, Z. & Jungers, R. M. Scenario-based set invariance verification for black-box nonlinear systems. IEEE Control. Syst. Lett. 5, 193–198 (2020)

    work page 2020

  66. [66]

    Dembo, R. S. Scenario optimization. Annals Oper. Res. 30, 63–80 (1991)

    work page 1991

  67. [67]

    & Rossi, F

    Bernardeschi, C., Lettieri, G. & Rossi, F. Statistical model checking of cooperative autonomous driving systems. In International Symposium on Leveraging Applications of Formal Methods, 316–332 (Springer, 2024)

    work page 2024

  68. [68]

    & Dullerud, G

    Wang, Y ., Roohi, N., West, M., Viswanathan, M. & Dullerud, G. E. Statistical verification of PCTL using antithetic and stratified samples. Formal Methods Syst. Des. 54, 145–163 (2019)

    work page 2019

  69. [69]

    Chow, Y . S. & Robbins, H. On the asymptotic theory of fixed-width sequential confidence intervals for the mean.The Annals Math. Stat. 36, 457–462 (1965)

    work page 1965

  70. [70]

    Clopper, C. J. & Pearson, E. S. The use of confidence or fiducial limits illustrated in the case of the binomial. Biometrika 26, 404–413 (1934)

    work page 1934

  71. [71]

    & Tsitsiklis, J

    Bertsekas, D. & Tsitsiklis, J. N. Neuro-dynamic programming (Athena Scientific, 1996). 13

    work page 1996

  72. [72]

    J., Katz, S

    Kochenderfer, M. J., Katz, S. M., Corso, A. L. & Moss, R. J. Algorithms for Validation (MIT Press, forthcoming)

    work page

  73. [73]

    & Pappas, G

    Lindemann, L., Jiang, L., Matni, N. & Pappas, G. J. Risk of stochastic systems for temporal logic specifications. ACM Transactions on Embed. Comput. Syst. 22, 1–31 (2023)

    work page 2023

  74. [74]

    & Ames, A

    Akella, P., Ahmadi, M. & Ames, A. D. A scenario approach to risk-aware safety-critical system verification. arXiv preprint arXiv:2203.02595 (2022)

    work page arXiv 2022

  75. [75]

    & Ferreira, A

    Haan, L. & Ferreira, A. Extreme value theory: An introduction, vol. 3 (Springer, 2006)

    work page 2006

  76. [76]

    Jenkinson, A. F. The frequency distribution of the annual maximum (or minimum) values of meteorological elements. Q. J. Royal meteorological society 81, 158–171 (1955)

    work page 1955

  77. [77]

    L., Shin, B

    Allen, B. L., Shin, B. T. & Cooper, P. J. Analysis of traffic conflicts and collision. InTransportation Research Record, 67–74 (1978)

    work page 1978

  78. [78]

    & Tarko, A

    Songchitruksa, P. & Tarko, A. P. The extreme value theory approach to safety estimation. Accid. Analysis & Prev. 38, 811–822 (2006)

    work page 2006

  79. [79]

    & Seshia, S

    Dreossi, T., Donzé, A. & Seshia, S. A. Compositional falsification of cyber-physical systems with machine learning components. J. Autom. Reason. 63, 1031–1053 (2019)

    work page 2019

  80. [80]

    How well generative adversarial networks learn distributions

    Liang, T. How well generative adversarial networks learn distributions. J. Mach. Learn. Res. 22, 1–41 (2021)

    work page 2021

Showing first 80 references.