Research map / Reinforcement learning / MDPs and planning
POMDPs and confounded RL
Study RL under partial observability and confounding.
99 accepted papers at ICML, NeurIPS, ICLR, CVPR and AAAI (2017–2026) · an approach in MDPs and planning, keyword cluster “markov decision · processes · observable”
State History Tried Checklist Preprints Who Export
Where it stands
Not enough benchmark data Fewer than four papers report the same benchmark, so there is no record to follow.
33 papers in 2023–24 and 21 in 2025–26 (at ICML, NeurIPS, ICLR, CVPR and AAAI); its share of the topic shrank, from 16.3% to 12.0%.
How this is decided
A benchmark is tracked when at least four papers of this approach report their own method on the same dataset and metric in at least two different years. Its record is the best number reported up to 2024 compared with the best reported in 2025–26: still improving means at least 1% better, flat means at most 0.2% better with at least two papers from 2025–26 trying. Numbers on a different scale from the rest (a fraction among percentages) and far outliers are left out. Results come from different papers with different setups, so they are signals, not head-to-head comparisons. The share of papers is counted at ICML, NeurIPS, ICLR, CVPR and AAAI.
How it got here
Most cited paper of each year
- 2017 · Unifying Task Specification in Reinforcement Learning (ICML 2017 · 17 citations)
- 2018 · A Lyapunov-based Approach to Safe Reinforcement Learning (NeurIPS 2018 · 77 citations)
- 2019 · Classification with Costly Features Using Deep Reinforcement Learning (AAAI 2019 · 73 citations)
- 2020 · Mitigating Bias in Face Recognition Using Skewness-Aware Reinforcement Learning (CVPR 2020 · 229 citations)
- 2021 · Iterative Bounding MDPs: Learning Interpretable Policies via Non-Interpretable Methods (AAAI 2021 · 20 citations)
- 2022 · Saute RL: Almost Surely Safe Reinforcement Learning Using State Augmentation (ICML 2022 · 9 citations)
- 2023 · Provably Efficient Causal Model-Based Reinforcement Learning for Systematic Generalization (AAAI 2023 · 7 citations)
- 2024 · Long-Term Safe Reinforcement Learning with Binary Feedback (AAAI 2024 · 4 citations)
- 2025 · Probabilistic Shielding for Safe Reinforcement Learning (AAAI 2025 · 5 citations)
- 2026 · Learn to change the world: Multi-level reinforcement learning with model-changing actions (ICML 2026 · 0 citations)
Earliest papers here
- Unifying Task Specification in Reinforcement Learning (ICML 2017 · 17 citations)
- #Exploration: A Study of Count-Based Exploration for Deep Reinforcement Learning (NeurIPS 2017)
- A Lyapunov-based Approach to Safe Reinforcement Learning (NeurIPS 2018 · 77 citations)
What has been tried
84 papers here report 582 measured changes (a component added, removed or swapped, compared with a baseline): helped 66%, no clear effect 10%, hurt 17%, mixed 7%. Papers mostly publish changes that helped, so "helped" is over-represented.
Ablations in recent papers
| What was changed | Measured on | Result | Effect | Paper |
|---|---|---|---|---|
| Use Feedback Alignment instead of regular backpropagation for recurrent gradient propagation | many environments · reward | Mixed | Real-Time Recurrent Reinforcement Learning (AAAI 2025 · 4 citations) | |
| Add GRSN gated recurrent neurons to LIF under TAP | Pendulum-P · mean reward | -824.8 -> -195.8 | Helped | GRSN: Gated Recurrent Spiking Neurons for POMDPs and MARL (AAAI 2025 · 1 citation) |
| replace next-latent-state prediction with next-observation prediction | 20 MiniGrid tasks · episode return | Helped | Bridging State and History Representations: Understanding Self-Predictive RL (ICLR 2024 · 1 citation) | |
| Add weighted averaging across equivalence classes during inference | Pollutant diffusion · estimation performance | Helped | Geometric Active Exploration in Markov Decision Processes: the Benefit of Abstraction (ICML 2024 · 0 citations) | |
| Use pessimistic rather than optimistic Q-function initialization | Tabular MDPs and Mon-MDPs · greedy policy return | Hurt | Beyond Optimism: Exploration With Partially Observable Rewards (NeurIPS 2024 · 0 citations) | |
| add a sequential encoder to quantized encoding | Synthetic Case 1 · PCC | 0.646 -> 0.910 | Helped | Identifying Latent State-Transition Processes for Individualized Reinforcement Learning (NeurIPS 2024 · 0 citations) |
| Use a lower learning rate for the context encoder while keeping the other layers at 3e-4 | Classic POMDP tasks · final return | Helped | Efficient Recurrent Off-Policy RL Requires a Context-Encoder-Specific Learning Rate (NeurIPS 2024 · 0 citations) |
What reviewers will expect
Datasets and baselines named in the 13 papers of this approach from 2025–2026 that have a full-text summary. Leaving out the common ones is a frequent review question.
Baselines
- PPO · 2 of 13 papers
Recent preprints
Loading recent arXiv preprints…
Who publishes on it
Institutions on the most papers of this approach since 2023 (of 54 papers; from OpenAlex and the papers' affiliations).
- ETH Zurich · 4 papers
- Carnegie Mellon University · 4 papers
- Meta · 3 papers
- Politecnico di Milano · 3 papers
- University of Oxford · 3 papers
- University of California, Berkeley · 3 papers
- ETHZ - ETH Zurich · 3 papers
- Massachusetts Institute of Technology · 3 papers
- Google · 3 papers
- McGill University · 2 papers
Use it
Download BibTeX of 40 key papersOpen the topic in the interactive map
Get alerts. Describe your idea in scime atlas and save it (free scime account): you get a weekly email when new preprints land close to it.
Related approaches
Same keyword cluster
- Average-reward and regularized MDPs (135 papers)
- Robust MDP policy gradient (87 papers)
- POMDP policy optimization (68 papers)
Elsewhere in the topic
- Function approximation in RL theory (93 papers)
- Offline RL sample complexity (86 papers)
- Policy optimization theory (61 papers)
- Rich-observation exploration (53 papers)
Benchmark numbers, ablations and baseline numbers are extracted by AI from each paper's full text and can be wrong; the paper is the authority. Data: results index of 2026-11-02, citation counts from OpenAlex (2026-10-09).