ICML 2022 accepted papers on Reinforcement learning
125 of the 1,233 papers accepted at ICML 2022 (10.1%) are on Reinforcement learning, compared with 11.7% at ICML 2021. They fall into 5 topics.
See ICML 2022 in the interactive map
Submitting to ICML? Checklist →
Working on something in this area? Describe your idea in scime atlas to see the closest papers by meaning and how crowded the spot has become.
Topics and papers
Policy gradient and actor-critic · 39 papers
- Temporal Difference Learning for Model Predictive Control (26 citations)
- Constrained Variational Policy Optimization for Safe Reinforcement Learning (16 citations)
- Off-Policy Reinforcement Learning with Delayed Rewards (14 citations)
- The Primacy Bias in Deep Reinforcement Learning (14 citations)
- Scalable Deep Reinforcement Learning Algorithms for Mean Field Games (13 citations)
- Policy Gradient Method For Robust Reinforcement Learning (11 citations)
- Mirror Learning: A Unifying Framework of Policy Optimisation (6 citations)
- Reachability Constrained Reinforcement Learning (6 citations)
- Lyapunov Density Models: Constraining Distribution Shift in Learning-Based Control (5 citations)
- Direct Behavior Specification via Constrained Reinforcement Learning (4 citations)
- Optimistic Linear Support and Successor Features as a Basis for Optimal Policy Transfer (4 citations)
- Reducing Variance in Temporal-Difference Value Estimation via Ensemble of Deep Networks (4 citations)
and 27 more in the interactive map.
Visual and representation RL · 33 papers
- Goal Misgeneralization in Deep Reinforcement Learning (21 citations)
- Evolving Curricula with Regret-Based Environment Design (19 citations)
- Recurrent Model-Free RL Can Be a Strong Baseline for Many POMDPs (17 citations)
- Reinforcement Learning with Action-Free Pre-Training from Videos (14 citations)
- Stabilizing Off-Policy Deep Reinforcement Learning from Pixels (8 citations)
- Transformers are Meta-Reinforcement Learners (8 citations)
- Causal Dynamics Learning for Task-Independent State Abstraction (6 citations)
- Retrieval-Augmented Reinforcement Learning (6 citations)
- Action-Sufficient State Representation Learning for Control with Structural Constraints (5 citations)
- Bisimulation Makes Analogies in Goal-Conditioned Reinforcement Learning (5 citations)
- Denoised MDPs: Learning World Models Better Than the World Itself (5 citations)
- Improving Policy Optimization with Generalist-Specialist Learning (5 citations)
and 21 more in the interactive map.
MDPs and planning · 21 papers
- Guarantees for Epsilon-Greedy Reinforcement Learning with Function Approximation (26 citations)
- Saute RL: Almost Surely Safe Reinforcement Learning Using State Augmentation (9 citations)
- Optimizing Sequential Experimental Design with Deep Reinforcement Learning (8 citations)
- Pessimistic Q-Learning for Offline Reinforcement Learning: Towards Optimal Sample Complexity (8 citations)
- Independent Policy Gradient for Large-Scale Markov Potential Games: Sharper Rates, Function Approximation, and Game-Agnostic Convergence (7 citations)
- Reinforcement Learning from Partial Observation: Linear Function Approximation with Provable Sample Efficiency (5 citations)
- Reward-Free RL is No Harder Than Reward-Aware RL in Linear Markov Decision Processes (5 citations)
- A Simple Reward-free Approach to Constrained Reinforcement Learning (4 citations)
- Contrastive UCB: Provably Efficient Contrastive Self-Supervised Learning in Online Reinforcement Learning (3 citations)
- Pessimistic Minimax Value Iteration: Provably Efficient Equilibrium Learning from Offline Datasets (3 citations)
- Flow-based Recurrent Belief State Learning for POMDPs (2 citations)
- Human-in-the-loop: Provably Efficient Preference-based Reinforcement Learning with General Function Approximation (2 citations)
and 9 more in the interactive map.
Offline RL · 18 papers
- Online Decision Transformer (21 citations)
- Prompting Decision Transformer for Few-Shot Policy Generalization (18 citations)
- Robust Task Representations for Offline Meta-Reinforcement Learning via Contrastive Learning (10 citations)
- Offline Meta-Reinforcement Learning with Online Self-Supervision (8 citations)
- On the Role of Discount Factor in Offline Reinforcement Learning (8 citations)
- Adversarially Trained Actor Critic for Offline Reinforcement Learning (7 citations)
- Addressing Optimism Bias in Sequence Modeling for Reinforcement Learning (5 citations)
- How to Leverage Unlabeled Data in Offline Reinforcement Learning (5 citations)
- Offline RL Policies Should Be Trained to be Adaptive (4 citations)
- Phasic Self-Imitative Reduction for Sparse-Reward Goal-Conditioned Reinforcement Learning (2 citations)
- Regularizing a Model-based Policy Stationary Distribution to Stabilize Offline Reinforcement Learning (2 citations)
- Koopman Q-learning: Offline Reinforcement Learning via Symmetries of Dynamics (1 citation)
and 6 more in the interactive map.
Imitation learning · 14 papers
- Discriminator-Weighted Offline Imitation Learning from Suboptimal Demonstrations (9 citations)
- Align-RUDDER: Learning From Few Demonstrations by Reward Redistribution (8 citations)
- Imitation Learning by Estimating Expertise of Demonstrators (7 citations)
- Causal Imitation Learning under Temporally Correlated Noise (5 citations)
- Modeling Strong and Human-Like Gameplay with KL-Regularized Search (4 citations)
- Robust Imitation Learning against Variations in Environment Dynamics (3 citations)
- A Hierarchical Bayesian Approach to Inverse Reinforcement Learning with Symbolic Reward Machines (2 citations)
- Continuous Control with Action Quantization from Demonstrations (2 citations)
- Plan Your Target and Learn Your Skills: Transferable State-Only Imitation Learning via Decoupled Policy Optimization (2 citations)
- Versatile Offline Imitation from Observations and Examples via Regularized State-Occupancy Matching (2 citations)
- Delayed Reinforcement Learning by Imitation (1 citation)
- Balancing Sample Efficiency and Suboptimality in Inverse Reinforcement Learning
and 2 more in the interactive map.
