Research map / Reinforcement learning / Offline RL
COptiDICE: Offline Constrained Reinforcement Learning via Stationary Distribution Correction Estimation
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 COptiDICE instead of Baseline or C-SPIBB | Offline constrained RL on random tabular CMDPs with constraint-satisfying data · Random tabular CMDPs | constraint satisfaction | Helped |
| Use COptiDICE instead of Baseline or C-SPIBB | Offline constrained RL on random tabular CMDPs with constraint-satisfying data · Random tabular CMDPs | normalized reward | Helped |
| Use COptiDICE instead of the baseline algorithms | Offline constrained RL on random tabular CMDPs with constraint-violating data · Random tabular CMDPs | constraint satisfaction | Helped |
| Use COptiDICE instead of BC | Offline constrained RL on RWRL continuous-control tasks · RWRL Suite | reward return | Helped |
| Use COptiDICE instead of the other evaluated algorithms | Offline constrained RL on RWRL continuous-control tasks · RWRL Suite | constraint satisfaction | Helped |
| Replace the upper-bound cost estimate with the vanilla OPE cost estimate (ablation) | COptiDICE (naive variant) · Random tabular CMDPs | constraint satisfaction | Hurt |
| Replace the upper-bound cost estimate with the vanilla OPE cost estimate (ablation) | COptiDICE (naive variant) · RWRL Suite | constraint satisfaction | Hurt |
| Increase the cost-bound uncertainty parameter epsilon (ablation) | COptiDICE · RWRL Suite | cost conservativeness | Helped |
| Increase the proportion of high-reward, high-cost trajectories in the dataset | Offline RL agents · RWRL Walker dataset configurations | reward return | Helped |
| Increase the proportion of high-reward, high-cost trajectories in the dataset | Offline RL agents · RWRL Walker dataset configurations | cost | Hurt |
| Use C-CRR instead of BC | Constrained offline RL on RWRL tasks · RWRL Suite | reward return | Hurt |
| Use CRR instead of BC | Offline RL on RWRL tasks · RWRL Suite | constraint satisfaction | Hurt |
Closest papers by meaning
- Relaxed Stationary Distribution Correction Estimation for Improved Offline Policy Optimization (AAAI 2024 · similarity 0.90)
- A Primal-Dual Algorithm for Offline Constrained Reinforcement Learning with Linear MDPs (ICML 2024 · similarity 0.90)
- Selective Uncertainty Propagation in Offline RL (AAAI 2025 · similarity 0.90)
- POCE: Primal Policy Optimization with Conservative Estimation for Multi-constraint Offline Reinforcement Learning (CVPR 2024 · similarity 0.90)
- Online Optimization for Offline Safe Reinforcement Learning (NeurIPS 2025 · similarity 0.89)
- ComaDICE: Offline Cooperative Multi-Agent Reinforcement Learning with Stationary Distribution Shift Regularization (ICLR 2025 · similarity 0.89)
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.