Pith. sign in

REVIEW 3 major objections 4 minor 1 cited by

Predictive Red Teaming: Breaking Policies Without Breaking Robots

T0 review · 3 major / 4 minor · reviewed 2026-08-08 · deepseek-v4-flash

Pith's one-line read The paper claims that a policy's success rate under an environmental change equals one minus the fraction of edited images the policy flags as anomalous, a relation it validates on two visuomotor diffusion policies across twelve…

desk verdict Good problem formulation and a real empirical baseline for hardware-free red teaming, but the edit-to-real fidelity question deserves a direct test before the quantitative claim is taken at face value. read the letter →

arxiv 2502.06575 v1 pith:QLDHDN4O submitted 2025-02-10 cs.RO cs.AIcs.LGcs.SYeess.SY

classification cs.ROcs.AIcs.LGcs.SYeess.SY
keywords predictiveredteaminganomalydetectiongenerativeimageeditingvisuomotorpoliciesdiffusionconformalpredictionrobotmanipulationtargeteddatacollection
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 proposes predictive red teaming: discovering which environmental changes (lighting, distractors, object or table position) will hurt a robot policy, and predicting how much, without running the policy on hardware in those conditions. It claims that a policy-specific anomaly rate over generative-image-edited observations predicts the real success rate under each change, and demonstrates this for two visuomotor diffusion policies across twelve off-nominal conditions with 500+ hardware trials. The reported accuracy is an average difference between predicted and real success rates below 0.19, with factor rankings by predicted degradation strongly correlated with real rankings. If true, the result would make vulnerability discovery for manipulation policies inexpensive and automated, and would enable targeted data collection that improves performance by 2-7x in adverse conditions.

What carries the argument

The load-bearing object is the anomaly rate $\alpha_f^\pi$ defined through the score $s^\pi(o, S_{\text{nom}})$, the mean of the $k$ smallest cosine distances between the embedding $\phi^\pi(o)$ of an observation and embeddings of nominal observations. Edited observations come from the Imagen 3 language-conditioned diffusion model, filtered by a vision-language model that keeps only edits faithfully reflecting the prompt; the threshold $\tau$ is the empirical quantile of nominal validation scores chosen by conformal prediction so that the nominal anomaly rate matches $1 - R^\pi_{\text{nom}}$. The identity $R_f^\pi \approx 1 - \alpha_f^\pi$ is the mechanism that turns image editing and embedding distances into a hardware-free performance forecast.

What would settle it

Take a factor with a known edit-to-real gap, such as lighting: real lighting changes alter shadows while the paper's edits leave shadows fixed. Run the full RoboART pipeline to predict success under that lighting from edited images, then measure the anomaly rate on real observations recorded under the same lighting; if the two anomaly rates differ by more than the reported prediction error, the identity $R_f^\pi \approx 1 - \alpha_f^\pi$ is measuring edit artifacts rather than the environmental factor. A broader test is to apply the pipeline to a purely non-visual change, such as object mass, where edited images are unchanged; the method predicts no degradation while hardware would show one, which would falsify the identity outside the visually-editable regime.

Watch

Extended reading notes

Core claim

The paper's central claim is the identity $R_f^\pi \approx 1 - \alpha_f^\pi$: the expected success of policy $\pi$ under an environmental factor $f$ equals one minus the anomaly rate $\alpha_f^\pi$, the fraction of edited observations flagged as out-of-distribution by a policy-embedding anomaly detector. The anomaly detector uses the mean $k$-nearest-neighbor cosine distance between an edited observation's latent representation (the policy's context vector) and nominal training observations, with a conformal-prediction threshold calibrated to the policy's nominal success rate. The paper validates this identity on two policies, a hybrid architecture combining trajectory optimization with diffusion heads and a vanilla diffusion policy, over twelve factors including three lighting colors, three background colors, four distractor objects, a person, and table height. It reports a Spearman rank correlation of 0.8 and 0.7 for the two policies and average absolute prediction errors of 0.10 and 0.19, and shows that fine-tuning with data from the three conditions predicted to be worst lifts performance 2-7x there and 2-5x in related conditions.

Load-bearing premise

The assumption that holds everything up is that an edited image shows the policy the same thing it would see in the real changed scenario, so the anomaly rate on edited images tracks the anomaly rate the policy would actually encounter.

Editorial extensions

If this is right

  • Robots can be screened for vulnerability to lighting, distractors, and placement changes using only image edits and embedding distances, no hardware runs in the off-nominal conditions.
  • The predicted factor rankings let a deployment team choose where a policy is safe to run and which of several policies generalizes better.
  • Data collection can be steered to the predicted worst conditions, producing 2-7x gains in those conditions and 2-5x in related ones.
  • Because the calibration uses conformal prediction, the false-positive rate on nominal observations is bounded by the user-chosen nominal anomaly rate.

Reading between the lines

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

  • If the identity holds across more policies and environments, predictive red teaming could be run continuously during development, flagging regressions after every training run without a hardware test.
  • The same edit-and-predict loop could be applied to vision-language-action foundation models, where the embedding is read from the policy's own transformer layers.
  • The reliance on first-frame observations suggests a testable extension: anomaly rates computed over short observation sequences might close part of the anomaly-to-failure gap the paper reports.
  • When edited images and real observations diverge, the pipeline's output should be read as an upper bound on robustness rather than a certified forecast; a user could build a calibration set of edit-vs-real pairs to quantify that bound.
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

3 major / 4 minor

Summary. The paper formalizes predictive red teaming for visuomotor policies: given nominal observations and a policy, discover environmental factors that degrade performance and predict the absolute or relative performance drop without running hardware evaluations in those scenarios. The proposed method, RoboART, first generates edited observations for each factor using a language-conditioned image editor (Imagen 3) filtered by a VLM critic, then computes a policy-embedding k-NN anomaly rate alpha for the edited observations and predicts the success rate as R_pred = 1 - alpha, with an anomaly threshold calibrated by conformal prediction to the known nominal success rate. The central claim, Eq. (1), is that R_f^pi is well approximated by 1 - alpha_f^pi. The paper evaluates this on two visuomotor diffusion policies across twelve off-nominal factors with 500+ hardware trials, reporting Spearman correlations of 0.8 and 0.7 and average absolute prediction errors of 0.10 and 0.19 for the two policies, and further shows that targeted data collection guided by the predictions improves performance by 2-7x in the predicted adverse conditions.

Significance. If the relation R_f^pi = 1 - alpha_f^pi holds on edited observations as a faithful proxy for real off-nominal conditions, the paper would provide a practical, hardware-free method for vulnerability discovery, policy comparison, and targeted data collection, which is valuable for deployment of imitation-learned manipulation policies. The empirical evaluation is substantial: 500+ hardware trials, two architecturally distinct policies, twelve factors, ablations over k and |S_nom|, and a demonstration of utility via fine-tuning. The paper also honestly lists limitations, including the edit-to-real gap and multi-view inconsistency. The main experimental contributions are useful even if the central equality is only an empirical approximation. However, the load-bearing claim that edited observations are faithful proxies for real off-nominal observations from the policy's perspective is not directly tested, and this gap is the key reason the manuscript needs revision rather than acceptance in its current form.

major comments (3)
  1. [Sec. 5.3, Eq. (1), Algorithm 1] The paper never compares the anomaly rate alpha_f^pi computed on edited observations S_f with the anomaly rate alpha_f,real^pi computed on real observations S_real_f, even though both sets exist for the same twelve factors. This comparison is the direct test of the edit-to-real link that separates Eq. (1) as a statement about the real off-nominal distribution from a statement about image-edit artifacts. Please include a per-factor comparison of alpha_f^pi and alpha_f,real^pi for both policies (a scatter plot or table with Spearman correlation and mean absolute difference), and discuss which factors diverge and why. Without this analysis, observed agreement between edited-based predictions and realized success could be driven by edit artifacts rather than by the environmental factor itself.
  2. [Appendix E, Table 1] For pi_dfn, the conformal threshold was recomputed using a nominal success rate of 0.8 instead of the measured 0.65 because the 0.65-calibrated threshold flagged most off-nominal real observations as anomalous. This is an additional calibration knob that directly affects the pi_dfn column of Table 1, but the paper does not state whether Table 1 and Fig. 5 use the recalibrated threshold. Please state which threshold was used for the main results and provide a sensitivity analysis showing the predictions under both thresholds. Otherwise the reported average error of 0.19 is not a parameter-free verification of Eq. (1).
  3. [Appendix A, 'Table height' edit; Fig. 5] The table-height factor is generated by changing the mat color and then applying a zoom to the image, which is not a faithful geometric proxy for a physical table-height change: the edit does not alter the overhead camera's perspective or the wrist camera's view consistently, and it cannot reproduce the change in reach that the real scenario introduces. Since table height is one of the factors with the largest predicted degradation, the claim that RoboART discovers the real vulnerability relies heavily on this edit being faithful. Please either use a geometrically consistent edit for this factor or provide per-factor evidence that the zoom-induced anomaly scores align with the anomaly scores from real table-height observations.
minor comments (4)
  1. [Fig. 5 caption] The caption states that the bottom panel is for "pi_hyb (vanilla diffusion policy)"; this should presumably be pi_dfn.
  2. [Abstract, Table 1] The abstract claims "less than 0.19 average difference," but Table 1 reports exactly 0.19 for pi_dfn; the wording should be "at most 0.19" or should report the value with the appropriate inequality.
  3. [Appendix E] The phrase "5x fewer observations" is confusing; a clearer formulation is "one fifth as many observations."
  4. [Table 1 and Table 2] Given that the Spearman correlations are computed with n = 12 factors, reporting p-values or 95% confidence intervals would help the reader calibrate the strength of the ranking evidence.

Circularity Check

0 steps flagged · score 0.0 of 10

No significant circularity: the nominal calibration is an input, off-nominal predictions are not fitted to hardware outcomes, and the anomaly-to-success link is an explicitly stated empirical hypothesis.

full rationale

RoboART's derivation chain is: (i) generate edited observations S_f with a third-party image editor; (ii) compute a policy-embedding anomaly score; (iii) choose the anomaly threshold tau so that alpha_nom = 1 - R_nom; (iv) predict R_f,pred = 1 - alpha_f for each off-nominal factor. Step (iii) is a calibration on nominal performance, which is given as an input; the paper never presents R_nom as a prediction. The twelve off-nominal success rates are not used to fit tau, the embedding, k, S_nom, or the editing prompts, so the reported Spearman correlations and average prediction errors are not forced by construction. Equation (1) is explicitly called the 'primary hypothesis' rather than derived from Equation (4), and Section 5.3 independently tests the anomaly-to-success link on real observations, which is exactly the kind of external check that would be absent if the central claim reduced to a definition. The paper's self-citations (e.g., [27], [31], [46], [60]) appear in related-work context and are not load-bearing; no uniqueness theorem is imported from the authors' prior work. The limitations discussed in Section 6.1 — the edit-to-real gap, multi-view inconsistency, and the approximation involved in the table-height edit — are empirical validity threats about whether S_f faithfully represents D_f, not circularities in the derivation. Overall, the prediction chain is self-contained: the only equation made true by construction is the nominal calibration equality, which is not counted as a predicted result.

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

No invented physical entities. The method's predictions rest on several free hyperparameters and on the assumption that synthetic edited images and policy embeddings behave like real off-nominal data.

free parameters (3)
  • k (number of nearest neighbors) = 5 for πhyb, 10 for πdfn
    Chosen based on ablations in Appendix D; affects anomaly scores and hence predicted degradation.
  • |S_nom| (nominal reference set size) = 3000 for πhyb, 500 for πdfn
    Selected via ablations in Appendix D; smaller sets help the high-dimensional πdfn embedding.
  • Nominal success rate used for πdfn conformal threshold = 0.8 (vs measured 0.65)
    Post hoc recalibration in Appendix E to avoid flagging all off-nominal observations as anomalous.
assumptions (5)
  • domain assumption Policy embedding distance is a sufficient proxy for task performance under distribution shift.
    The entire method relies on the anomaly-to-failure gap being small; the paper acknowledges this as a limitation (Sec 6.1).
  • domain assumption Generative edits faithfully reflect the specified environmental factor in policy-relevant ways.
    Imagen 3 edits are assumed to be realistic and to preserve policy-relevant structure; the paper acknowledges the edit-to-real gap (Sec 6.1) and the table-height edit is only an approximation (Appendix A).
  • domain assumption Nominal observations S_nom ∪ S_val are representative of nominal deployment.
    The conformal threshold is calibrated on these observations; if the nominal training distribution is not representative, anomaly rates will be miscalibrated (Sec 4.2).
  • standard math Conformal prediction guarantees apply to the chosen threshold.
    The threshold in Eq. 4 uses conformal quantiles; the paper invokes [67] for the probability bound.
  • domain assumption The VLM critic reliably selects edits that adhere to the prompt.
    If the VLM selects unfaithful edits, S_f contains noise that distorts anomaly rates (Sec 4.1).

how reviews work

0 comments
Cite this review

Pith. "Pith review of Predictive Red Teaming: Breaking Policies Without Breaking Robots." pith.science (2026). https://pith.science/paper/QLDHDN4O

@misc{pith2026250206575,
  author       = {Pith},
  title        = {Pith review of: Predictive Red Teaming: Breaking Policies Without Breaking Robots},
  year         = {2026},
  howpublished = {\url{https://pith.science/paper/QLDHDN4O}},
  note         = {Machine review of arXiv:2502.06575}
}
read the original abstract

Visuomotor policies trained via imitation learning are capable of performing challenging manipulation tasks, but are often extremely brittle to lighting, visual distractors, and object locations. These vulnerabilities can depend unpredictably on the specifics of training, and are challenging to expose without time-consuming and expensive hardware evaluations. We propose the problem of predictive red teaming: discovering vulnerabilities of a policy with respect to environmental factors, and predicting the corresponding performance degradation without hardware evaluations in off-nominal scenarios. In order to achieve this, we develop RoboART: an automated red teaming (ART) pipeline that (1) modifies nominal observations using generative image editing to vary different environmental factors, and (2) predicts performance under each variation using a policy-specific anomaly detector executed on edited observations. Experiments across 500+ hardware trials in twelve off-nominal conditions for visuomotor diffusion policies demonstrate that RoboART predicts performance degradation with high accuracy (less than 0.19 average difference between predicted and real success rates). We also demonstrate how predictive red teaming enables targeted data collection: fine-tuning with data collected under conditions predicted to be adverse boosts baseline performance by 2-7x.

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. SAFECAST: Robust Failure Detection for VLA Policies with Contrast-Set Training and Calibration

    cs.RO 2026-08 conditional novelty 5.0 of 10

    SAFECAST augments hidden-state failure-probe training and conformal calibration with visual and language contrast sets, improving VLA failure detection under distribution shift in several tested settings.

Reference graph

Works this paper leans on

79 extracted references · 38 canonical work pages · cited by 1 Pith paper

  1. [1]

    Diffusion policy: Visuomotor policy learning via action diffusion

    Cheng Chi, Zhenjia Xu, Siyuan Feng, Eric Cousineau, Yilun Du, Benjamin Burchfiel, Russ Tedrake, and Shuran Song. Diffusion policy: Visuomotor policy learning via action diffusion. The International Journal of Robotics Research, 2023

  2. [2]

    Imagen 3.arXiv preprint arXiv:2408.07009, 2024

    Jason Baldridge, Jakob Bauer, Mukul Bhutani, Nicole Brichtova, Andrew Bunner, Kelvin Chan, Yichang Chen, Sander Dieleman, Yuqing Du, Zach Eaton-Rosen, et al. Imagen 3.arXiv preprint arXiv:2408.07009, 2024

  3. [3]

    Scaling rectified flow transformers for high-resolution image synthesis

    Patrick Esser, Sumith Kulal, Andreas Blattmann, Rahim Entezari, Jonas Müller, Harry Saini, Yam Levi, Dominik Lorenz, Axel Sauer, Frederic Boesel, et al. Scaling rectified flow transformers for high-resolution image synthesis. InProceedings of the International Conference on Machine Learning, 2024

  4. [4]

    Adding conditional control to text-to-image diffusion models

    Lvmin Zhang, Anyi Rao, and Maneesh Agrawala. Adding conditional control to text-to-image diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 3836–3847, 2023

  5. [5]

    Photoreal- istic text-to-image diffusion models with deep language understanding.Advances in neural information processing systems, 35:36479–36494, 2022

    Chitwan Saharia, William Chan, Saurabh Saxena, Lala Li, Jay Whang, Emily L Denton, Kamyar Ghasemipour, Raphael Gontijo Lopes, Burcu Karagol Ayan, Tim Salimans, et al. Photoreal- istic text-to-image diffusion models with deep language understanding.Advances in neural information processing systems, 35:36479–36494, 2022

  6. [6]

    Springer, 2005

    Vladimir Vovk, Alexander Gammerman, and Glenn Shafer.Algorithmic Learning in a Random World, volume 29. Springer, 2005. 13 Predictive Red Teaming: Breaking Policies Without Breaking Robots

  7. [7]

    Basic Books, 2015

    Micah Zenko.Red Team: How To Succeed By Thinking Like The Enemy. Basic Books, 2015

  8. [8]

    GPT-4 technical report.arXiv preprint arXiv:2303.08774, 2023

    Josh Achiam, Steven Adler, Sandhini Agarwal, Lama Ahmad, Ilge Akkaya, Florencia Leoni Aleman, Diogo Almeida, Janko Altenschmidt, Sam Altman, Shyamal Anadkat, et al. GPT-4 technical report.arXiv preprint arXiv:2303.08774, 2023

Show all 79 references
  1. [9]

    Gemini: a family of highly capable multimodal models.arXiv preprint arXiv:2312.11805, 2023

    Gemini Team, Rohan Anil, Sebastian Borgeaud, Jean-Baptiste Alayrac, Jiahui Yu, Radu Soricut, Johan Schalkwyk, Andrew M Dai, Anja Hauth, Katie Millican, et al. Gemini: a family of highly capable multimodal models.arXiv preprint arXiv:2312.11805, 2023

  2. [10]

    Red teaming language models to reduceharms: Methods, scalingbehaviors, andlessonslearned

    Deep Ganguli, Liane Lovitt, Jackson Kernion, Amanda Askell, Yuntao Bai, Saurav Kadavath, Ben Mann, Ethan Perez, Nicholas Schiefer, Kamal Ndousse, et al. Red teaming language models to reduceharms: Methods, scalingbehaviors, andlessonslearned. arXivpreprintarXiv:2209.07858 , 2022

  3. [11]

    Training a helpful and harmless assistant with reinforcement learning from human feedback.arXiv preprint arXiv:2204.05862, 2022

    Yuntao Bai, Andy Jones, Kamal Ndousse, Amanda Askell, Anna Chen, Nova DasSarma, Dawn Drain, Stanislav Fort, Deep Ganguli, Tom Henighan, et al. Training a helpful and harmless assistant with reinforcement learning from human feedback.arXiv preprint arXiv:2204.05862, 2022

  4. [12]

    Jailbroken: How does LLM safety training fail? Advances in Neural Information Processing Systems, 36, 2024

    Alexander Wei, Nika Haghtalab, and Jacob Steinhardt. Jailbroken: How does LLM safety training fail? Advances in Neural Information Processing Systems, 36, 2024

  5. [13]

    Red teaming language models with language models

    Ethan Perez, Saffron Huang, Francis Song, Trevor Cai, Roman Ring, John Aslanides, Amelia Glaese, Nat McAleese, and Geoffrey Irving. Red teaming language models with language models. arXiv preprint arXiv:2202.03286, 2022

  6. [14]

    Mass-producing failures of multimodal systems with language models.Advances in Neural Information Processing Systems, 36, 2024

    Shengbang Tong, Erik Jones, and Jacob Steinhardt. Mass-producing failures of multimodal systems with language models.Advances in Neural Information Processing Systems, 36, 2024

  7. [15]

    Universal and transferable adversarial attacks on aligned language models.arXiv preprint arXiv:2307.15043, 2023

    Andy Zou, Zifan Wang, Nicholas Carlini, Milad Nasr, J Zico Kolter, and Matt Fredrikson. Universal and transferable adversarial attacks on aligned language models.arXiv preprint arXiv:2307.15043, 2023

  8. [16]

    Jailbreaking black box large language models in twenty queries.arXiv preprint arXiv:2310.08419, 2023

    Patrick Chao, Alexander Robey, Edgar Dobriban, Hamed Hassani, George J Pappas, and Eric Wong. Jailbreaking black box large language models in twenty queries.arXiv preprint arXiv:2310.08419, 2023

  9. [17]

    AutoDAN: Generating stealthy jailbreak prompts on aligned large language models.arXiv preprint arXiv:2310.04451, 2023

    Xiaogeng Liu, Nan Xu, Muhao Chen, and Chaowei Xiao. AutoDAN: Generating stealthy jailbreak prompts on aligned large language models.arXiv preprint arXiv:2310.04451, 2023

  10. [18]

    Tree of attacks: Jailbreaking black-box LLMs automatically.arXiv preprint arXiv:2312.02119, 2023

    Anay Mehrotra, Manolis Zampetakis, Paul Kassianik, Blaine Nelson, Hyrum Anderson, Yaron Singer, and Amin Karbasi. Tree of attacks: Jailbreaking black-box LLMs automatically.arXiv preprint arXiv:2312.02119, 2023

  11. [19]

    Red teaming visual language models.arXiv preprint arXiv:2401.12915, 2024

    Mukai Li, Lei Li, Yuwei Yin, Masood Ahmed, Zhenguang Liu, and Qi Liu. Red teaming visual language models.arXiv preprint arXiv:2401.12915, 2024

  12. [20]

    Arondight: Red teaming large vision language models with auto-generated multi-modal jailbreak prompts

    Yi Liu, Chengjun Cai, Xiaoli Zhang, Xingliang Yuan, and Cong Wang. Arondight: Red teaming large vision language models with auto-generated multi-modal jailbreak prompts. InProceedings of the 32nd ACM International Conference on Multimedia, pages 3578–3586, 2024

  13. [21]

    Red-teaming the stable diffusion safety filter.arXiv preprint arXiv:2210.04610, 2022

    Javier Rando, Daniel Paleka, David Lindner, Lennart Heim, and Florian Tramèr. Red-teaming the stable diffusion safety filter.arXiv preprint arXiv:2210.04610, 2022. 14 Predictive Red Teaming: Breaking Policies Without Breaking Robots

  14. [22]

    Erasing concepts from diffusion models

    Rohit Gandikota, Joanna Materzynska, Jaden Fiotto-Kaufman, and David Bau. Erasing concepts from diffusion models. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 2426–2436, 2023

  15. [23]

    Embodied red teaming for auditing robotic foundation models.arXiv preprint arXiv:2411.18676, 2024

    Sathwik Karnik, Zhang-Wei Hong, Nishant Abhangi, Yen-Chen Lin, Tsun-Hsuan Wang, and Pulkit Agrawal. Embodied red teaming for auditing robotic foundation models.arXiv preprint arXiv:2411.18676, 2024

  16. [24]

    Jailbreaking LLM-controlled robots.arXiv preprint arXiv:2410.13691, 2024

    Alexander Robey, Zachary Ravichandran, Vijay Kumar, Hamed Hassani, and George J Pappas. Jailbreaking LLM-controlled robots.arXiv preprint arXiv:2410.13691, 2024

  17. [25]

    THE COLOSSEUM: A benchmark for evaluating generalization for robotic manipulation

    Wilbert Pumacay, Ishika Singh, Jiafei Duan, Ranjay Krishna, Jesse Thomason, and Dieter Fox. THE COLOSSEUM: A benchmark for evaluating generalization for robotic manipulation. 2024

  18. [26]

    Reachability-based safe learning with gaussian processes

    Anayo K Akametalu, Jaime F Fisac, Jeremy H Gillula, Shahab Kaynama, Melanie N Zeilinger, and Claire J Tomlin. Reachability-based safe learning with gaussian processes. InProceedings of the 53rd IEEE Conference on Decision and Control, pages 1424–1431, 2014

  19. [27]

    Sim-to-lab- to-real: Safe reinforcement learning with shielding and generalization guarantees.Artificial Intelligence, 314:103811, 2023

    Kai-Chieh Hsu, Allen Z Ren, Duy P Nguyen, Anirudha Majumdar, and Jaime F Fisac. Sim-to-lab- to-real: Safe reinforcement learning with shielding and generalization guarantees.Artificial Intelligence, 314:103811, 2023

  20. [28]

    The safety filter: A unified view of safety-critical control in autonomous systems.Annual Review of Control, Robotics, and Autonomous Systems, 7, 2023

    Kai-Chieh Hsu, Haimin Hu, and Jaime F Fisac. The safety filter: A unified view of safety-critical control in autonomous systems.Annual Review of Control, Robotics, and Autonomous Systems, 7, 2023

  21. [29]

    Control barrier function based quadratic programs for safety critical systems.IEEE Transactions on Automatic Control, 62(8): 3861–3876, 2016

    Aaron D Ames, Xiangru Xu, Jessy W Grizzle, and Paulo Tabuada. Control barrier function based quadratic programs for safety critical systems.IEEE Transactions on Automatic Control, 62(8): 3861–3876, 2016

  22. [30]

    Safe reinforcement learning via shielding

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

  23. [31]

    Ren, and Anirudha Majumdar

    Alec Farid, David Snyder, Allen Z. Ren, and Anirudha Majumdar. Failure prediction with statistical guarantees for vision-based robot control. InProceedings of Robotics: Science and Systems (RSS), 2022

  24. [32]

    When to ask for help: Proactive interventions in autonomous reinforcement learning.Advances in Neural Information Processing Systems, 35:16918–16930, 2022

    Annie Xie, Fahim Tajwar, Archit Sharma, and Chelsea Finn. When to ask for help: Proactive interventions in autonomous reinforcement learning.Advances in Neural Information Processing Systems, 35:16918–16930, 2022

  25. [33]

    Asking for help: Failure prediction in behavioral cloning through value approximation

    Cem Gokmen, Daniel Ho, and Mohi Khansari. Asking for help: Failure prediction in behavioral cloning through value approximation. InProceedings of the IEEE International Conference on Robotics and Automation (ICRA), pages 5821–5828, 2023

  26. [34]

    Model-basedruntimemonitoring with interactive imitation learning

    HuihanLiu,ShivinDass,RobertoMartín-Martín,andYukeZhu. Model-basedruntimemonitoring with interactive imitation learning. InProceedings of the IEEE International Conference on Robotics and Automation (ICRA), pages 4154–4161, 2024

  27. [35]

    Safe visual navigation via deep learning and novelty detection

    Charles Richter and Nicholas Roy. Safe visual navigation via deep learning and novelty detection. In Proceedings of Robotics: Science and Systems (RSS), 2017. 15 Predictive Red Teaming: Breaking Policies Without Breaking Robots

  28. [36]

    A system-level view on out-of-distribution data in robotics.arXiv preprint arXiv:2212.14020, 2022

    Rohan Sinha, Apoorva Sharma, Somrita Banerjee, Thomas Lew, Rachel Luo, Spencer M Richards, Yixiao Sun, Edward Schmerling, and Marco Pavone. A system-level view on out-of-distribution data in robotics.arXiv preprint arXiv:2212.14020, 2022

  29. [37]

    A unified survey on anomaly, novelty, open-set, and out- of-distribution detection: Solutions and future challenges.arXiv preprint arXiv:2110.14051, 2021

    Mohammadreza Salehi, Hossein Mirzaei, Dan Hendrycks, Yixuan Li, Mohammad Hossein Rohban, and Mohammad Sabokrou. A unified survey on anomaly, novelty, open-set, and out- of-distribution detection: Solutions and future challenges.arXiv preprint arXiv:2110.14051, 2021

  30. [38]

    Real-time anomaly detection and reactive planning with large language models

    Rohan Sinha, Amine Elhafsi, Christopher Agia, Matthew Foutter, Edward Schmerling, and Marco Pavone. Real-time anomaly detection and reactive planning with large language models. arXiv preprint arXiv:2407.08735, 2024

  31. [39]

    Unsupervised anomaly detection for self-flying delivery drones

    Vikas Sindhwani, Hakim Sidahmed, Krzysztof Choromanski, and Brandon Jones. Unsupervised anomaly detection for self-flying delivery drones. In2020 IEEE international conference on robotics and automation (ICRA), pages 186–192. IEEE, 2020

  32. [40]

    Online learning and sequential anomaly detection in trajectories

    Rikard Laxhammar and Göran Falkman. Online learning and sequential anomaly detection in trajectories. IEEE Transactions on Pattern Analysis and Machine Intelligence, 36(6):1158–1173, 2013

  33. [41]

    Sample-efficient safety assurances using conformal prediction.The International Journal of Robotics Research, 43(9):1409–1424, 2024

    Rachel Luo, Shengjia Zhao, Jonathan Kuck, Boris Ivanovic, Silvio Savarese, Edward Schmer- ling, and Marco Pavone. Sample-efficient safety assurances using conformal prediction.The International Journal of Robotics Research, 43(9):1409–1424, 2024

  34. [42]

    Closing the loop on runtime monitors with fallback-safe MPC

    Rohan Sinha, Edward Schmerling, and Marco Pavone. Closing the loop on runtime monitors with fallback-safe MPC. In2023 62nd IEEE Conference on Decision and Control (CDC), pages 6533–6540. IEEE, 2023

  35. [43]

    Safe planning in dynamic environments using conformal prediction.IEEE Robotics and Automation Letters, 2023

    Lars Lindemann, Matthew Cleaveland, Gihyun Shim, and George J Pappas. Safe planning in dynamic environments using conformal prediction.IEEE Robotics and Automation Letters, 2023

  36. [44]

    Adaptive conformal prediction for motion planning among dynamic agents

    Anushri Dixit, Lars Lindemann, Skylar X Wei, Matthew Cleaveland, George J Pappas, and Joel W Burdick. Adaptive conformal prediction for motion planning among dynamic agents. InLearning for Dynamics and Control Conference, pages 300–314. PMLR, 2023

  37. [45]

    Robots that ask for help: Uncertainty alignment for large language model planners.arXiv preprint arXiv:2307.01928, 2023

    Allen Z Ren, Anushri Dixit, Alexandra Bodrova, Sumeet Singh, Stephen Tu, Noah Brown, Peng Xu, Leila Takayama, Fei Xia, Jake Varley, et al. Robots that ask for help: Uncertainty alignment for large language model planners.arXiv preprint arXiv:2307.01928, 2023

  38. [46]

    Perceive with confidence: Statistical safety assurances for navigation with learning- based perception

    AnushriDixit, ZhitingMei, MeghanBooker, MarikoStorey-Matsutani, AllenZRen, andAnirudha Majumdar. Perceive with confidence: Statistical safety assurances for navigation with learning- based perception. InProceedings of the Conference on Robot Learning (CoRL), 2024

  39. [47]

    Formal verification and control with conformal prediction.arXiv preprint arXiv:2409.00536, 2024

    Lars Lindemann, Yiqi Zhao, Xinyi Yu, George J Pappas, and Jyotirmoy V Deshmukh. Formal verification and control with conformal prediction.arXiv preprint arXiv:2409.00536, 2024

  40. [48]

    Improving image generation with better captions.OpenAI, 2023

    James Betker, Gabriel Goh, Li Jing, Tim Brooks, Jianfeng Wang, Linjie Li, Long Ouyang, Juntang Zhuang, Joyce Lee, Yufei Guo, et al. Improving image generation with better captions.OpenAI, 2023

  41. [49]

    Glide: Towards photorealistic image generation and editing with text-guided diffusion models.arXiv preprint arXiv:2112.10741, 2021

    Alex Nichol, Prafulla Dhariwal, Aditya Ramesh, Pranav Shyam, Pamela Mishkin, Bob McGrew, Ilya Sutskever, and Mark Chen. Glide: Towards photorealistic image generation and editing with text-guided diffusion models.arXiv preprint arXiv:2112.10741, 2021. 16 Predictive Red Teaming...

  42. [50]

    Inpaint anything: Segment anything meets image inpainting.arXiv preprint arXiv:2304.06790, 2023

    Tao Yu, Runseng Feng, Ruoyu Feng, Jinming Liu, Xin Jin, Wenjun Zeng, and Zhibo Chen. Inpaint anything: Segment anything meets image inpainting.arXiv preprint arXiv:2304.06790, 2023

  43. [51]

    Editgan: High-precision semantic image editing.Advances in Neural Information Processing Systems, 34:16331–16345, 2021

    Huan Ling, Karsten Kreis, Daiqing Li, Seung Wook Kim, Antonio Torralba, and Sanja Fidler. Editgan: High-precision semantic image editing.Advances in Neural Information Processing Systems, 34:16331–16345, 2021

  44. [52]

    In-domain GAN inversion for real image editing

    Jiapeng Zhu, Yujun Shen, Deli Zhao, and Bolei Zhou. In-domain GAN inversion for real image editing. InEuropean conference on computer vision, pages 592–608. Springer, 2020

  45. [53]

    Genaug: Retargeting behaviors to unseen situations via generative augmentation.arXiv preprint arXiv:2302.06671, 2023

    Zoey Chen, Sho Kiami, Abhishek Gupta, and Vikash Kumar. Genaug: Retargeting behaviors to unseen situations via generative augmentation.arXiv preprint arXiv:2302.06671, 2023

  46. [54]

    Scaling robot learning with semantically imagined experience.arXiv preprint arXiv:2302.11550, 2023

    Tianhe Yu, Ted Xiao, Austin Stone, Jonathan Tompson, Anthony Brohan, Su Wang, Jaspiar Singh, Clayton Tan, Jodilyn Peralta, Brian Ichter, et al. Scaling robot learning with semantically imagined experience.arXiv preprint arXiv:2302.11550, 2023

  47. [55]

    Roboagent: Generalization and efficiency in robot manipulation via semantic augmentations and action chunking

    Homanga Bharadhwaj, Jay Vakil, Mohit Sharma, Abhinav Gupta, Shubham Tulsiani, and Vikash Kumar. Roboagent: Generalization and efficiency in robot manipulation via semantic augmentations and action chunking. InProceedings of the IEEE International Conference on Robotics and Aut...

  48. [56]

    RoVi- Aug: Robot and viewpoint augmentation for cross-embodiment robot learning.arXiv preprint arXiv:2409.03403, 2024

    Lawrence Yunliang Chen, Chenfeng Xu, Karthik Dharmarajan, Muhammad Zubair Irshad, Richard Cheng, Kurt Keutzer, Masayoshi Tomizuka, Quan Vuong, and Ken Goldberg. RoVi- Aug: Robot and viewpoint augmentation for cross-embodiment robot learning.arXiv preprint arXiv:2409.03403, 2024

  49. [57]

    Semantically controllable augmentations for generalizable robot learning

    Zoey Chen, Zhao Mandi, Homanga Bharadhwaj, Mohit Sharma, Shuran Song, Abhishek Gupta, and Vikash Kumar. Semantically controllable augmentations for generalizable robot learning. The International Journal of Robotics Research, page 02783649241273686, 2024

  50. [58]

    Zero-shot robotic manipulation with pretrained image-editing diffusion models

    Kevin Black, Mitsuhiko Nakamoto, Pranav Atreya, Homer Walke, Chelsea Finn, Aviral Kumar, and Sergey Levine. Zero-shot robotic manipulation with pretrained image-editing diffusion models. arXiv preprint arXiv:2310.10639, 2023

  51. [59]

    Vint: A foundation model for visual navigation.arXiv preprint arXiv:2306.14846, 2023

    Dhruv Shah, Ajay Sridhar, Nitish Dashora, Kyle Stachowicz, Kevin Black, Noriaki Hirose, and Sergey Levine. Vint: A foundation model for visual navigation.arXiv preprint arXiv:2306.14846, 2023

  52. [60]

    Run-time observation interventions make vision-language-action models more visually robust.arXiv preprint arXiv:2410.01971, 2024

    Asher J Hancock, Allen Z Ren, and Anirudha Majumdar. Run-time observation interventions make vision-language-action models more visually robust.arXiv preprint arXiv:2410.01971, 2024

  53. [61]

    Eligibility traces for off-policy policy evaluation.Computer Science Department Faculty Publication Series, page 80, 2000

    Doina Precup. Eligibility traces for off-policy policy evaluation.Computer Science Department Faculty Publication Series, page 80, 2000

  54. [62]

    Consistent on-line off-policy evaluation

    Assaf Hallak and Shie Mannor. Consistent on-line off-policy evaluation. InProceedings of the International Conference on Machine Learning (ICML), pages 1372–1383, 2017

  55. [63]

    Bootstrapping with models: Confidence intervals foroff-policyevaluation

    Josiah Hanna, Peter Stone, and Scott Niekum. Bootstrapping with models: Confidence intervals foroff-policyevaluation. In ProceedingsoftheAAAIConferenceonArtificialIntelligence , volume31, 2017. 17 Predictive Red Teaming: Breaking Policies Without Breaking Robots

  56. [64]

    More robust doubly robust off-policy evaluation

    Mehrdad Farajtabar, Yinlam Chow, and Mohammad Ghavamzadeh. More robust doubly robust off-policy evaluation. InProceedings of the International Conference on Machine Learning (ICML), pages 1447–1456, 2018

  57. [65]

    Offline reinforcement learning: Tutorial, review, and perspectives on open problems.arXiv preprint arXiv:2005.01643, 2020

    Sergey Levine, Aviral Kumar, George Tucker, and Justin Fu. Offline reinforcement learning: Tutorial, review, and perspectives on open problems.arXiv preprint arXiv:2005.01643, 2020

  58. [66]

    Online distribution shift detection via recency prediction

    Rachel Luo, Rohan Sinha, Yixiao Sun, Ali Hindy, Shengjia Zhao, Silvio Savarese, Edward Schmerling, and Marco Pavone. Online distribution shift detection via recency prediction. In 2024 IEEE International Conference on Robotics and Automation (ICRA), pages 16251–16263. IEEE, 2024

  59. [67]

    A gentle introduction to conformal prediction and distribution-free uncertainty quantification.arXiv preprint arXiv:2107.07511, 2021

    Anastasios N Angelopoulos and Stephen Bates. A gentle introduction to conformal prediction and distribution-free uncertainty quantification.arXiv preprint arXiv:2107.07511, 2021

  60. [68]

    Hydra: Hybrid robot actions for imitation learning

    Suneel Belkhale, Yuchen Cui, and Dorsa Sadigh. Hydra: Hybrid robot actions for imitation learning. InProceedings of the Conference on Robot Learning, pages 2113–2133, 2023

  61. [69]

    Spearman rank correlation.Encyclopedia of Biostatistics, 7, 2005

    Jerrold H Zar. Spearman rank correlation.Encyclopedia of Biostatistics, 7, 2005

  62. [70]

    Proedit: Simple progression is all you need for high-quality 3D scene editing.arXiv preprint arXiv:2411.05006, 2024

    Jun-Kun Chen and Yu-Xiong Wang. Proedit: Simple progression is all you need for high-quality 3D scene editing.arXiv preprint arXiv:2411.05006, 2024

  63. [71]

    3d Gaussian splatting: Survey, technologies, challenges, and opportunities.arXiv preprint arXiv:2407.17418, 2024

    Yanqi Bao, Tianyu Ding, Jing Huo, Yaoli Liu, Yuxin Li, Wenbin Li, Yang Gao, and Jiebo Luo. 3d Gaussian splatting: Survey, technologies, challenges, and opportunities.arXiv preprint arXiv:2407.17418, 2024

  64. [72]

    Uncertainty-aware failure detection for imitation learning robot policies

    Chen Xu, Tony Khuong Nguyen, Patrick Miller, Robert Lee, Paarth Shah, Rares Andrei Ambrus, Haruki Nishimura, and Masha Itkina. Uncertainty-aware failure detection for imitation learning robot policies. InCoRL Workshop on Safe and Robust Robot Learning for Operation in the Real World

  65. [73]

    An image is worth 16x16 words: Transformers for image recognition at scale

    Alexey Dosovitskiy, Lucas Beyer, Alexander Kolesnikov, Dirk Weissenborn, Xiaohua Zhai, Thomas Unterthiner, Mostafa Dehghani, Matthias Minderer, Georg Heigold, Sylvain Gelly, Jakob Uszkor- eit, and Neil Houlsby. An image is worth 16x16 words: Transformers for image recognition ...

  66. [74]

    Token- learner: Adaptive space-time tokenization for videos.Advances in neural information processing systems, 34:12786–12797, 2021

    Michael Ryoo, AJ Piergiovanni, Anurag Arnab, Mostafa Dehghani, and Anelia Angelova. Token- learner: Adaptive space-time tokenization for videos.Advances in neural information processing systems, 34:12786–12797, 2021

  67. [75]

    Scalable diffusion models with transformers

    William Peebles and Saining Xie. Scalable diffusion models with transformers. InProceedings of the IEEE/CVF International Conference on Computer Vision, pages 4195–4205, 2023

  68. [76]

    Goal-conditioned imitation learning using score-based diffusion policies.arXiv preprint arXiv:2304.02532, 2023

    Moritz Reuss, Maximilian Li, Xiaogang Jia, and Rudolf Lioutikov. Goal-conditioned imitation learning using score-based diffusion policies.arXiv preprint arXiv:2304.02532, 2023

  69. [77]

    <short edit instruction >

    TeroKarras, MiikaAittala, TimoAila, andSamuliLaine. Elucidatingthedesignspaceofdiffusion- based generative models.Advances in neural information processing systems, 35:26565–26577, 2022. 18 Predictive Red Teaming: Breaking Policies Without Breaking Robots A. Image Editing: Exa...

  70. [78]

    a waypoint action mode which outputs a single waypoint(𝑤∈ 𝑆𝐸(3)), and

  71. [79]

    a trajectory action mode which outputs a dense sequence of robot joint angles(𝑞𝑖∈ ℝ14). In addition to these policy heads we also output a mode selection scalar which defines which action 21 Predictive Red Teaming: Breaking Policies Without Breaking Robots Image Encoder Image ...

Pith tools

Reviewed August 8, 2026 · model on record in the stance chip above.