The DOI matches its Crossref record; 6 of the 6 references it cites were found in a registry or an index.
Understanding RL vision
Open accessOverview
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Integrity signals
Registry
Cites
Statistics
Verification
Claims the paper makes
The claims and statistics of this paper haven’t been checked yet.
Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.
Claims other papers make about this one
Each is checked against this paper’s own text: does it say what the citing paper says it does?
Hilton et al. (2020) validated their methods by explaining cases where a deep reinforcement learning agent's neural network failed to achieve maximum reward and also explained specific hallucinations exhibited by the network.
“For instance, Hilton et al. (2020) validated their methods by explaining cases where a deep reinforcement learning agent's neural network failed to achieve maximum reward and also explained specific hallucinations exhibited by the network.”
From Open Problems in Mechanistic Interpretability (2025) · cites this paper as Hilton et al., 2020
The “Dissecting failure” and “Hallucinations” sections describe explaining failures to achieve maximum reward and specific mistaken detections, including a background shape mistaken for a coin. But the “Model editing” section says the authors quantitatively validated their analysis by editing the model and testing the effects, not by explaining those cases. (LLM, conf=0.86)
Experimental — accuracy not yet measured
Self-check
Percentages, sums and reported differences re-computed from the paper’s own numbers.
The self-check needs the paper’s full text.
Citations
Cites
ImageNet: A large-scale hierarchical image database
Deng, J., Dong, W., Socher, R., Li, L.-J., Li, K., & Fei-Fei, L. (2009) · 2009 IEEE Conference on Computer Vision and Pattern Recognition · doi:10.1109/cvpr.2009.5206848 · Cited by 65,333
Proximal Policy Optimization Algorithms
Schulman, J., Wolski, F., Dhariwal, P., Radford, A., & Klimov, O. (2017) · arXiv · doi:10.48550/arxiv.1707.06347 · Cited by 10,625
Reconciling modern machine-learning practice and the classical bias–variance trade-off
Belkin, M., Hsu, D., Ma, S., & Mandal, S. (2019) · Proceedings of the National Academy of Sciences · doi:10.1073/pnas.1903070116 · Cited by 1,595
Dota 2 with Large Scale Deep Reinforcement Learning
OpenAI, :, Berner, C., Brockman, G., Chan, B., Cheung, V., Dębiak, P., Dennison, C., Farhi, D., Fischer, Q., et al. (2019) · arXiv · doi:10.48550/arxiv.1912.06680 · Cited by 1,065
Human-level performance in 3D multiplayer games with population-based reinforcement learning
Jaderberg, M., Czarnecki, W. M., Dunning, I., Marris, L., Lever, G., Castañeda, A. G., Beattie, C., Rabinowitz, N. C., Morcos, A. S., Ruderman, A., et al. (2019) · Science · doi:10.1126/science.aau6249 · Cited by 662
Observational Overfitting in Reinforcement Learning
Song, X., Jiang, Y., Tu, S., Du, Y., & Neyshabur, B. (2019) · arXiv · doi:10.48550/arxiv.1912.02975 · Cited by 48
Flagged first, then by citations
Research works printed but not linked: 31
- Cobbe, Quantifying generalization in reinforcement learning, arXiv preprint arXiv:1812.02341 · arXiv
- Simonyan, Deep inside convolutional networks: Visualising image classification models and saliency maps, arXiv preprint arXiv:1312.6034 · arXiv
- Zeiler, Visualizing and understanding convolutional networks, 10.1007/978-3-319-10590-1_53 · DOI · Paper page
- Springenberg, Striving for simplicity: The all convolutional net, arXiv preprint arXiv:1412.6806 · arXiv
- Selvaraju, Grad-CAM: Visual explanations from deep networks via gradient-based localization, 10.1109/ICCV.2017.74 · DOI · Paper page
- Fong, Interpretable explanations of black boxes by meaningful perturbation, 10.1109/ICCV.2017.371 · DOI · Paper page
- Kindermans, PatternNet and PatternLRP--Improving the interpretability of neural networks, stat, 1050
- Kindermans, The (un)reliability of saliency methods, 10.1007/978-3-030-28954-6_14 · DOI
- Sundararajan, Axiomatic attribution for deep networks
- Olah, The Building Blocks of Interpretability, Distill, 10.23915/distill.00010 · DOI · Paper page
- Schulman, High-dimensional continuous control using generalized advantage estimation, arXiv preprint arXiv:1506.02438 · arXiv
- Cammarata, Thread: Circuits, Distill, 10.23915/distill.00024 · DOI
- Perez-Liebana, General Video Game AI: A multi-track framework for evaluating agents, games and content generation algorithms, arXiv preprint arXiv:1802.10363 · arXiv
- Juliani, Obstacle Tower: A Generalization Challenge in Vision, Control, and Planning, arXiv preprint arXiv:1902.01378 · arXiv
- Olah, Feature Visualization, Distill, 10.23915/distill.00007 · DOI · Paper page
- Erhan, Visualizing higher-layer features of a deep network, University of Montreal, 1341
- Nguyen, Deep neural networks are easily fooled: High confidence predictions for unrecognizable images, 10.1109/CVPR.2015.7298640 · DOI · Paper page
- Mordvintsev, Inceptionism: Going deeper into neural networks, Google Research Blog
- Nguyen, Plug & play generative networks: Conditional iterative generation of images in latent space, 10.1109/CVPR.2017.374 · DOI
- Szegedy, Going deeper with convolutions, 10.1109/CVPR.2015.7298594 · DOI · Paper page
- Such, An Atari model zoo for analyzing, visualizing, and comparing deep reinforcement learning agents, arXiv preprint arXiv:1812.07069 · arXiv
- Rupprecht, Finding and Visualizing Weaknesses of Deep Reinforcement Learning Agents, arXiv preprint arXiv:1904.01318 · arXiv
- Cammerata, Caricatures, Distill draft
- Madry, Towards deep learning models resistant to adversarial attacks, arXiv preprint arXiv:1706.06083 · arXiv
- Szegedy, Intriguing properties of neural networks, arXiv preprint arXiv:1312.6199 · arXiv
- Greydanus, Visualizing and understanding Atari agents, arXiv preprint arXiv:1711.00138 · arXiv
- Ilyas, Adversarial examples are not bugs, they are features, arXiv preprint arXiv:1905.02175 · arXiv
- Engstrom, A Discussion of 'Adversarial Examples Are Not Bugs, They Are Features', Distill, 10.23915/distill.00019 · DOI
- Akkaya, Solving Rubik's Cube with a Robot Hand, arXiv preprint arXiv:1910.07113 · arXiv
- Olah, Does Attribution Make Sense?, Distill draft
- Espeholt, IMPALA: Scalable distributed deep-RL with importance weighted actor-learner architectures, arXiv preprint arXiv:1802.01561 · arXiv
Cited by
Sorted by citations
Open Problems in Mechanistic Interpretability
Sharkey, L., Chughtai, B., Batson, J., Lindsey, J., Wu, J., Bushnaq, L., Goldowsky-Dill, N., Heimersheim, S., Ortega, A., Bloom, J., et al. (2025) · arXiv · doi:10.48550/arxiv.2501.16496 · Cited by 291
Open Problems and Fundamental Limitations of Reinforcement Learning from Human Feedback
Casper, S., Davies, X., Shi, C., Gilbert, T. K., Scheurer, J., Rando, J., Freedman, R., Korbak, T., Lindner, D., Freire, P., et al. (2023) · arXiv · doi:10.48550/arxiv.2307.15217 · Cited by 101
“Shah et al. [2022], Di Langosco et al. [2022] and Hilton et al. [2020] study this type of failure in-depth.”
Background
Metadata
- Authors
- Hilton, J., Cammarata, N., Carter, S., Goh, G., & Olah, C.
- Affiliations
- OpenAI (United States) (US)
- Observable
- ORCID
- Hilton, Jacob
- Venue
- Distill
- Publisher
- Distill Working Group
- Published
- 2020-11-17
- Volume
- 5
- Issue
- 11
- ISSN
- 2476-0757 (ISSN-L)
- OpenAlex
- W3100003809
- Cited by
- 37
- References
- 6
- Citations by year
- 2026: 3 · 2025: 6 · 2024: 7 · 2023: 10 · 2022: 7 · 2021: 4
- FWCI
- 1.77
- Type
- journal article
- Language
- en
- Open access
- diamond
- doi.org
- Licence
- cc-by
- Locations
- Distill (publishedVersion, cc-by)
- Topics
- Physical Sciences › Computer Science › Artificial Intelligence › Reinforcement Learning in Robotics
- Physical Sciences › Computer Science › Artificial Intelligence › Domain Adaptation and Few-Shot Learning
- Physical Sciences › Computer Science › Computer Vision and Pattern Recognition › Advanced Neural Network Applications
- Keywords
- reinforcement learning, vision, visual perception, visual servoing
About these checks
Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.
Statistics and arithmetic are recomputed, not judged.
Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.
What counts as a claim ›How we measure confidence ›How we check this ›
Are you an author, or do you see a problem with this page? Request a review
