Research map / Reinforcement learning / MDPs and planning
Deciding What to Model: Value-Equivalent Sampling for Reinforcement Learning
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 |
|---|---|---|---|
| Use rate-distortion compression of the posterior MDP, allowing distortion threshold D | PSRL, producing VSRL · Finite-horizon episodic MDPs | Bayesian regret upper bound: \sqrt{\overline{\Gamma}K\mc{R}^{\Pi,\mc{V}}_1(D)} + 2KH\sqrt{D} | Mixed |
| Replace the general Bellman-distortion measure d_{Π,V} with optimal action-value distortion d_{Q*} (ablation) | VSRL · Tabular MDPs | Bayesian regret upper bound: \mc{O}\left(|\mc{S}|\sqrt{\overline{\Gamma}|\mc{A}|K}\right) -> \widetilde{\mc{O}}\left(\sqrt{\overline{\Gamma}|\mc{S}|| | Helped |
| Limit the information retained by the compressed model to agent capacity R | VSRL with d_{Π,V} · Finite-horizon episodic MDPs | Bayesian regret upper bound: \sqrt{\overline{\Gamma}KR} + 2KH\sqrt{\mc{D}^{\Pi,\mc{V}}_1(R)} | Mixed |
| Limit the information retained by the compressed model to agent capacity R | VSRL with d_{Q*} · Finite-horizon episodic MDPs | Bayesian regret upper bound: \sqrt{\overline{\Gamma}KR} + 2K(H+1)\sqrt{\mc{D}^{Q^\star}_1(R)} | Mixed |
Closest papers by meaning
- The Value Equivalence Principle for Model-Based Reinforcement Learning (NeurIPS 2020 · similarity 0.94)
- Inverting the Bellman Equation: From $Q$-Values to World Models (NeurIPS 2026 · similarity 0.90)
- Approximate Value Equivalence (NeurIPS 2022 · similarity 0.90)
- Calibrated Value-Aware Model Learning with Probabilistic Environment Models (ICML 2025 · similarity 0.89)
- The Edge-of-Reach Problem in Offline Model-Based Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Mitigating Partial Observability in Sequential Decision Processes via the Lambda Discrepancy (NeurIPS 2024 · similarity 0.88)
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.