mlsanityData to 7 Oct 2026

Provable Guarantees for Generative Behavior Cloning: Bridging Low-Level Stability and High-Level Behavior

arXiv 2307.14619 ↗First posted 27 Jul 2023cs.LG math.ST stat.MLcited by 33

Not on the map. Why?

Abstract (arXiv)

We propose a theoretical framework for studying behavior cloning of complex expert demonstrations using generative modeling. Our framework invokes low-level controllers - either learned or implicit in position-command control - to stabilize imitation around expert demonstrations. We show that with (a) a suitable low-level stability guarantee and (b) a powerful enough generative model as our imitation learner, pure supervised behavior cloning can generate trajectories matching the per-time step distribution of essentially arbitrary expert trajectories in an optimal transport cost. Our analysis relies on a stochastic continuity property of the learned policy we call "total variation continuity" (TVC). We then show that TVC can be ensured with minimal degradation of accuracy by combining a popular data-augmentation regimen with a novel algorithmic trick: adding augmentation noise at execution time. We instantiate our guarantees for policies parameterized by diffusion models and prove that if the learner accurately estimates the score of the (noise-augmented) expert policy, then the distribution of imitator trajectories is close to the demonstrator distribution in a natural optimal transport distance. Our analysis constructs intricate couplings between noise-augmented trajectories, a technique that may be of independent interest. We conclude by empirically validating our algorithmic recommendations, and discussing implications for future research directions for better behavior cloning with generative modeling.

Key facts. Cited by 33 papers in the map (data through 7 Oct 2026). Builds on 54 of its 77 references that are in the map. First posted on arXiv 27 Jul 2023.

Builds on

54 of 77 references.

References · 54

  1. 2017Attention Is All You Needcited by 59,932
  2. 2020Denoising Diffusion Probabilistic Modelscited by 20,851
  3. 2015U-Net: Convolutional Networks for Biomedical Image Segmentationcited by 19,604
  4. 2020Denoising Diffusion Implicit Modelscited by 9,655
  5. 2015Deep Unsupervised Learning using Nonequilibrium Thermodynamicscited by 7,005
  6. 2022Hierarchical Text-Conditional Image Generation with CLIP Latentscited by 6,539
  7. 2019Generative Modeling by Estimating Gradients of the Data Distributioncited by 4,067
  8. 2021Improved Denoising Diffusion Probabilistic Modelscited by 3,736
  9. 2023Diffusion Policy: Visuomotor Policy Learning via Action Diffusioncited by 2,730
  10. 2017FiLM: Visual Reasoning with a General Conditioning Layercited by 2,286
  11. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,842
  12. 2021Decision Transformer: Reinforcement Learning via Sequence Modelingcited by 1,528
38 more
  1. 2018Group Normalizationcited by 1,514
  2. 2023Learning Fine-Grained Bimanual Manipulation with Low-Cost Hardwarecited by 1,490
  3. 2022LaMDA: Language Models for Dialog Applicationscited by 1,218
  4. 2022Planning with Diffusion for Flexible Behavior Synthesiscited by 1,023
  5. 2016End to End Learning for Self-Driving Carscited by 1,000
  6. 2021What Matters in Learning from Offline Human Demonstrations for Robot Manipulationcited by 699
  7. 2020Image Augmentation Is All You Need: Regularizing Deep Reinforcement Learning from Pixelscited by 573
  8. 2022Is Conditional Generative Modeling all you need for Decision-Making?cited by 504
  9. 2022Sampling is as easy as learning the score: theory for diffusion models with minimal data assumptionscited by 440
  10. 2018ChauffeurNet: Learning to Drive by Imitating the Best and Synthesizing the Worstcited by 391
  11. 2019Sliced Score Matching: A Scalable Approach to Density and Score Estimationcited by 381
  12. 2016YouTube-8M: A Large-Scale Video Classification Benchmarkcited by 324
  13. 2023IDQL: Implicit Q-Learning as an Actor-Critic Method with Diffusion Policiescited by 307
  14. 2022Behavior Transformers: Cloning k modes with one stonecited by 299
  15. 2017Deep Imitation Learning for Complex Manipulation Tasks from Virtual Reality Teleoperationcited by 286
  16. 2019Causal Confusion in Imitation Learningcited by 286
  17. 2019RoboNet: Large-Scale Multi-Robot Learningcited by 268
  18. 2023Imitating Human Behaviour with Diffusion Modelscited by 260
  19. 2018HG-DAgger: Interactive Imitation Learning with Human Expertscited by 245
  20. 2017Non-convex learning via Stochastic Gradient Langevin Dynamics: a nonasymptotic analysiscited by 242
  21. 2021Deep learning: a statistical viewpointcited by 239
  22. 2017One-Shot Visual Imitation Learning via Meta-Learningcited by 225
  23. 2017DART: Noise Injection for Robust Imitation Learningcited by 187
  24. 2022Convergence of score-based generative modeling for general data distributionscited by 179
  25. 2016A vector-contraction inequality for Rademacher complexitiescited by 141
  26. 2020Generative Modeling with Denoising Auto-Encoders and Langevin Samplingcited by 118
  27. 2019A Short Note on Concentration Inequalities for Random Vectors with SubGaussian Normcited by 105
  28. 2022Privacy of Noisy Stochastic Gradient Descent: More Iterations without More Privacy Losscited by 60
  29. 2023Faster high-accuracy log-concave sampling via algorithmic warm startscited by 44
  30. 2021On the Sample Complexity of Stability Constrained Imitation Learningcited by 38
  31. 2020Grasping with Chopsticks: Combating Covariate Shift in Model-free Imitation Learning for Fine Manipulationcited by 37
  32. 2021Mitigating Covariate Shift in Imitation Learning via Offline Data Without Great Coveragecited by 26
  33. 2013Empirical entropy, minimax regret and minimax riskcited by 21
  34. 2021Shortest Paths in Graphs of Convex Setscited by 21
  35. 2022TaSIL: Taylor Series Imitation Learningcited by 21
  36. 2023MEGA-DAgger: Imitation Learning with Multiple Imperfect Expertscited by 17
  37. 2020Fast Mixing of Multi-Scale Langevin Dynamics under the Manifold Hypothesiscited by 11
  38. 2022Distributionally Adaptive Meta Reinforcement Learningcited by 11

50 most cited of 54 All 54 (JSON)

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 3; 2024: 12; 2025: 14; 2026: 4.

33 papers.

Sort:

By area:

  1. 2024Diffusion Forcing: Next-token Prediction Meets Full-Sequence Diffusioncited by 634
  2. 2024Diffusion Policy Policy Optimizationcited by 255
  3. 2023Compositional Foundation Models for Hierarchical Planningcited by 92
  4. 2024PoCo: Policy Composition from and for Heterogeneous Robot Learningcited by 42
  5. 2024JUICER: Data-Efficient Imitation Learning for Robotic Assemblycited by 36
  6. 2025Much Ado About Noising: Dispelling the Myths of Generative Robotic Controlcited by 35
  7. 2025Streaming Flow Policy: Simplifying diffusion/flow-matching policies by treating action trajectories as flow trajectoriescited by 34
  8. 2023CCIL: Continuity-based Data Augmentation for Corrective Imitation Learningcited by 26
  9. 2025A Theory of Learning with Autoregressive Chain of Thoughtcited by 20
  10. 2024Diffusing States and Matching Scores: A New Framework for Imitation Learningcited by 16
  11. 2025One-Step Generative Policies with Q-Learning: A Reformulation of MeanFlowcited by 16
  12. 2023Robot Fleet Learning via Policy Mergingcited by 11
21 more
  1. 2024Active Fine-Tuning of Multi-Task Policiescited by 8
  2. 2024Hybrid Reinforcement Learning from Offline Observation Alonecited by 5
  3. 2025Learning with Imperfect Models: When Multi-step Prediction Mitigates Compounding Errorcited by 5
  4. 2025Inverse Q-Learning Done Right: Offline Imitation Learning in Qπ-Realizable MDPscited by 5
  5. 2024The Role of Inherent Bellman Error in Offline Reinforcement Learning with Linear Function Approximationcited by 4
  6. 2025Accelerating Diffusion Planners in Offline RL via Reward-Aware Consistency Trajectory Distillationcited by 3
  7. 2025How Well do Diffusion Policies Learn Kinematic Constraint Manifolds?cited by 3
  8. 2025To Distill or Decide? Understanding the Algorithmic Trade-off in Partially Observable Reinforcement Learningcited by 3
  9. 2026TMRL: Diffusion Timestep-Modulated Pretraining Enables Exploration for Efficient Policy Finetuningcited by 3
  10. 2025EMA Without the Lag: Bias-Corrected Iterate Averaging Schemescited by 2
  11. 2024Dense Dynamics-Aware Reward Synthesis: Integrating Prior Experience with Demonstrationscited by 1
  12. 2025A Test-Function Approach to Incremental Stabilitycited by 1
  13. 2025Continuous Vision-Language-Action Co-Learning with Semantic-Physical Alignment for Behavioral Cloningcited by 1
  14. 2025GrOMP: Grasped Object Manifold Projection for Multimodal Imitation Learning of Manipulationcited by 1
  15. 2026Flow as Flow: Modeling Robot Velocity Fields as Probability Velocity Fields for Flow-Based Object Manipulationcited by 1
  16. 2024Learning in Hybrid Active Inference Modelscited by 0
  17. 2024Improved Sample Complexity of Imitation Learning for Barrier Model Predictive Controlcited by 0
  18. 2024Problem Space Transformations for Out-of-Distribution Generalisation in Behavioural Cloningcited by 0
  19. 2025A Model-Based Approach to Imitation Learning through Multi-Step Predictionscited by 0
  20. 2026Understanding Behavior Cloning with Action Quantizationcited by 0
  21. 2026Behavior Cloning Under PD Control: A Finite-Horizon Theory of Gain-Dependent Error Amplificationcited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2024Stable-BC: Controlling Covariate Shift with Stable Behavior Cloning
  2. 2022Diverse Imitation Learning via Self-Organizing Generative Models
  3. 2025The Pitfalls of Imitation Learning when Actions are Continuous
  4. 2022Robust Imitation of a Few Demonstrations with a Backwards Model
  5. 2024Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning
  6. 2026Learning When to Stop: Selective Imitation Learning Under Arbitrary Dynamics Shift
  7. 2022Data augmentation for efficient learning from parametric experts
  8. 2024Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning