Research map / Reinforcement learning / Policy gradient and actor-critic
Scalable Safe Policy Improvement via Monte Carlo Tree Search
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 |
|---|---|---|---|
| MCTS-SPIBB (online, locally computed, with UCT merged with baseline bootstrapping) vs. full SPIBB policy iteration | MCTS-SPIBB · Gridworld and SysAdmin safety experiments | performance and 15%-CVaR relative to SPIBB | No clear effect |
| Replacing SPIBB with unsafe Basic RL (no baseline constraint) on small datasets | Basic RL · Gridworld |D|=2 and SysAdmin |D|=5 | performance relative to baseline | Hurt |
| Increasing MCTS simulations from 100 to 10000 when computing the policy (ablation) | MCTS-SPIBB · Gridworld 3x3/4x4/5x5 and SysAdmin 7 machines | absolute value difference vs. SPIBB in initial state (avg) | Helped |
| Using MCTS-SPIBB (online, local) instead of full-state SPIBB policy computation | SysAdmin scaling · SysAdmin up to 35 machines | maximum tractable machines | Helped |
| Using MCTS-SPIBB on WetChicken benchmark vs. other SPI algorithms | MCTS-SPIBB · WetChicken (Scholl et al. 2022b benchmark) | performance | No clear effect |
Closest papers by meaning
- Scalable Safe Policy Improvement for Factored Multi-Agent MDPs (ICML 2024 · similarity 0.89)
- Online Robust Reinforcement Learning Through Monte-Carlo Planning (ICML 2025 · similarity 0.88)
- Efficient Policy Evaluation with Offline Data Informed Behavior Policy Design (ICML 2024 · similarity 0.87)
- Trust-Region Twisted Policy Improvement (ICML 2025 · similarity 0.87)
- Long-Term Safe Reinforcement Learning with Binary Feedback (AAAI 2024 · similarity 0.86)
- Enhancing Efficiency of Safe Reinforcement Learning via Sample Manipulation (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.