Research map / Reinforcement learning / MDPs and planning
Provably efficient RL with Rich Observations via Latent State Decoding
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| latent-state decoding (PCID) with linear decoders instead of naïve epsilon-greedy exploration | PCID vs QLEARNING on combination lock · Lock-Bernoulli | time-to-solve vs horizon H: exponential improvement | Helped |
| PCID decoding with noisy Gaussian observations (sigma increased to 0.3) | PCID with linear functions · Lock-Gaussian stochastic, H=40 | solves within 100K episodes | Hurt |
| two-layer neural network decoder instead of linear decoder | PCID on Lock-Gaussian · Lock-Gaussian, sigma=0.3, H=30 | solves within 100K episodes | Hurt |
| underestimating clustering parameter k or sample size n (ablation) | PCID · combination lock sensitivity analysis | algorithm success | Hurt |
| overestimating clustering parameter k and sample size n (ablation) | PCID · combination lock sensitivity analysis | algorithm success | No clear effect |
Closest papers by meaning
- Provably Efficient Exploration for Reinforcement Learning Using Unsupervised Learning (NeurIPS 2020 · similarity 0.90)
- Provably Filtering Exogenous Distractors using Multistep Inverse Dynamics (ICLR 2022 · similarity 0.88)
- Restoring Exploration after Post-Training: Latent Exploration Decoding for Large Reasoning Models (ICML 2026 · similarity 0.87)
- Provable Rich Observation Reinforcement Learning with Combinatorial Latent States (ICLR 2021 · similarity 0.87)
- Rich-Observation Reinforcement Learning with Continuous Latent Dynamics (ICML 2024 · similarity 0.86)
- RL in Latent MDPs is Tractable: Online Guarantees via Off-Policy Evaluation (NeurIPS 2024 · similarity 0.86)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.