Research map / Reinforcement learning / Offline RL
Multi-Objective SPIBB: Seldonian Offline Policy Improvement with Safety Constraints in Finite MDPs
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 |
|---|---|---|---|
| Add individual-objective safety constraints to scalarized policy improvement (MO-SPIBB) | Offline policy improvement · Synthetic safety grid-world CMDPs | failure rate | Helped |
| Relax the Soft-SPIBB policy-deviation constraint by increasing ε (ablation) | MO-SPIBB · Synthetic safety grid-world CMDPs | failure rate: ε=0.1 -> ε=1.0 | Hurt |
| Use tuned ε for MO-SPIBB rather than fixed hyperparameters (ablation) | MO-SPIBB · Synthetic safety grid-world CMDPs | improvement over baseline | Helped |
| Use the extended empirical Bernstein bound instead of Student's t-test for HCPI safety testing (ablation) | MO-HCPI · Synthetic safety grid-world CMDPs | conservatism | Hurt |
| Remove rare state-action transitions from training rather than retaining them with a rarity penalty (ablation) | Offline sepsis treatment policy · MIMIC-III sepsis task | survival return: 65.95 -> 11.26 | Hurt |
Closest papers by meaning
- An Offline Adaptation Framework for Constrained Multi-Objective Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Constraint-Adaptive Policy Switching for Offline Safe Reinforcement Learning (AAAI 2025 · similarity 0.89)
- Online Optimization for Offline Safe Reinforcement Learning (NeurIPS 2025 · similarity 0.88)
- Latent Safety-Constrained Policy Approach for Safe Offline Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Safe Offline Reinforcement Learning with Feasibility-Guided Diffusion Model (ICLR 2024 · similarity 0.88)
- Safe and Efficient: A Primal-Dual Method for Offline Convex CMDPs under Partial Data Coverage (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.