mlsanityData to 7 Oct 2026

Robust Asymmetric Learning in POMDPs

arXiv 2012.15566 ↗First posted 31 Dec 2020cs.LG stat.MLcited by 23

Not on the map. Why?

Abstract (arXiv)

Policies for partially observed Markov decision processes can be efficiently learned by imitating policies for the corresponding fully observed Markov decision processes. Unfortunately, existing approaches for this kind of imitation learning have a serious flaw: the expert does not know what the trainee cannot see, and so may encourage actions that are sub-optimal, even unsafe, under partial information. We derive an objective to instead train the expert to maximize the expected reward of the imitating agent policy, and use it to construct an efficient algorithm, adaptive asymmetric DAgger (A2D), that jointly trains the expert and the agent. We show that A2D produces an expert policy that the agent can safely imitate, in turn outperforming policies learned by imitating a fixed expert.

Key facts. Cited by 23 papers in the map (data through 7 Oct 2026). First posted on arXiv 31 Dec 2020.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 3; 2022: 3; 2023: 4; 2024: 4; 2025: 8; 2026: 1.

23 papers.

Sort:

By area:

  1. 2021Learning Visible Connectivity Dynamics for Cloth Smoothingcited by 61
  2. 2022Hindsight Learning for MDPs with Exogenous Inputscited by 25
  3. 2025LEAD: Minimizing Learner-Expert Asymmetry in End-to-End Drivingcited by 22
  4. 2021GridToPix: Training Embodied Agents with Minimal Supervisioncited by 20
  5. 2024Experts Don't Cheat: Learning What You Don't Know By Predicting Pairscited by 18
  6. 2022Leveraging Fully Observable Policies for Learning under Partial Observabilitycited by 17
  7. 2023Learning in POMDPs is Sample-Efficient with Hindsight Observabilitycited by 17
  8. 2023TGRL: An Algorithm for Teacher Guided Reinforcement Learningcited by 16
  9. 2023Informed POMDP: Leveraging Additional Information in Model-Based RLcited by 11
  10. 2023Bridging the Sim-to-Real Gap from the Information Bottleneck Perspectivecited by 8
  11. 2022Improved Policy Optimization for Online Imitation Learningcited by 6
  12. 2024Student-Informed Teacher Trainingcited by 6
11 more
  1. 2024AgentMixer: Multi-Agent Correlated Policy Factorizationcited by 5
  2. 2025A Theoretical Justification for Asymmetric Actor-Critic Algorithmscited by 5
  3. 2021Seeing Differently, Acting Similarly: Heterogeneously Observable Imitation Learningcited by 4
  4. 2025Guided Policy Optimization under Partial Observabilitycited by 3
  5. 2025Multi-Agent Guided Policy Optimizationcited by 3
  6. 2025To Distill or Decide? Understanding the Algorithmic Trade-off in Partially Observable Reinforcement Learningcited by 3
  7. 2025Distilling Realizable Students from Unrealizable Teacherscited by 2
  8. 2025Informed Asymmetric Actor-Critic: Leveraging Privileged Signals Beyond Full-State Accesscited by 2
  9. 2024Sequential Decision Making with Expert Demonstrations under Unobserved Heterogeneitycited by 1
  10. 2025Communication-Enhanced Tutoring for Efficient Decentralized Multi-Agent Reinforcement Learningcited by 0
  11. 2026Reinformed Dreamer: An Asymmetric World Model Efficiently Trained through Latent Guidancecited by 0

Similar, not cited

Close in wording, no citation either way

  1. 2019Reinforced Imitation in Heterogeneous Action Space
  2. 2025The Pitfalls of Imitation Learning when Actions are Continuous
  3. 2025Imitation Learning as Return Distribution Matching
  4. 2025Learning safe, constrained policies via imitation learning: Connection to Probabilistic Inference and a Naive Algorithm
  5. 2021Provably Breaking the Quadratic Error Compounding Barrier in Imitation Learning, Optimally
  6. 2018Fast Policy Learning through Imitation and Reinforcement
  7. 2020Toward the Fundamental Limits of Imitation Learning
  8. 2022Imitation Learning by State-Only Distribution Matching