Research map / Reinforcement learning / Imitation learning
BAIL: Best-Action Imitation Learning for Batch Deep 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 |
|---|---|---|---|
| select best actions with upper envelope and imitate them, instead of BCQ | batch DRL on MuJoCo · 22 training batches | average return: 42% better than BCQ, 101% better than BC | Helped |
| use BAIL instead of BC/BCQ on execution batches | batch DRL on MuJoCo · 40 execution batches | average return | No clear effect |
| replace upper envelope with selecting top-return data points (ablation) | BAIL · MuJoCo batches | learning curves | Hurt |
| replace upper envelope with standard regression value function (ablation) | BAIL · MuJoCo batches | learning curves | Hurt |
| use augmentation heuristic instead of oracle 2000-step returns (ablation) | BAIL · Hopper-v2 | learning curves | No clear effect |
| train envelope and policy in parallel (Progressive BAIL) (ablation) | BAIL · MuJoCo batches | average return | No clear effect |
| use BAIL-difference instead of BAIL-ratio selection (ablation) | BAIL · MuJoCo batches | average return | No clear effect |
Closest papers by meaning
- Interactive and Hybrid Imitation Learning: Provably Beating Behavior Cloning (NeurIPS 2025 · similarity 0.86)
- Beyond-Expert Performance with Limited Demonstrations: Efficient Imitation Learning with Double Exploration (ICLR 2025 · similarity 0.86)
- Is Behavior Cloning All You Need? Understanding Horizon in Imitation Learning (NeurIPS 2024 · similarity 0.86)
- Hybrid Inverse Reinforcement Learning (ICML 2024 · similarity 0.86)
- Blending Imitation and Reinforcement Learning for Robust Policy Improvement (ICLR 2024 · similarity 0.86)
- Split the Differences, Pool the Rest: Provably Efficient Multi-Objective Imitation (NeurIPS 2026 · 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.