Research map / Reinforcement learning / Policy gradient and actor-critic
Planning and Learning with Adaptive Lookahead
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 adaptive state-dependent lookahead instead of a fixed horizon | TLPI and QLPI · Maze | simulator queries to convergence | Helped |
| Use TLPI with different tested contraction targets | TLPI · Maze | simulator queries to convergence | Helped |
| Use QLPI with tested quantile budgets | QLPI · Maze | simulator queries to convergence | Mixed |
| Replace exact V⋆ with an estimate from state aggregation (ablation) | QLPI · Maze | simulator queries to convergence | Hurt |
| Use state-aggregation estimates with different aggregation sizes (ablation) | QLPI · Maze | simulator queries to convergence | No clear effect |
| Use adaptive lookahead rather than a fixed tree depth | QL-DQN · VideoPinball | training reward over time | Helped |
| Use adaptive lookahead rather than a fixed tree depth | QL-DQN · Tutankham | training reward over time | Helped |
| Use adaptive lookahead rather than a fixed tree depth | QL-DQN · Solaris | training reward over time | No clear effect |
| Use adaptive lookahead rather than a fixed tree depth | QL-DQN · Berzerk | training reward over time | No clear effect |
| Use a separate value head for each search depth (ablation) | QL-DQN · Atari | training stability | Helped |
| Bootstrap one search depth using another depth's target (ablation) | QL-DQN · Atari | training stability | Hurt |
| Use fixed 1-step lookahead rather than the best fixed lookahead | h-PI · Maze | simulator queries to convergence | Hurt |
Closest papers by meaning
- Reinforcement Learning with Multi-Step Lookahead Information Via Adaptive Batching (NeurIPS 2026 · similarity 0.89)
- Reinforcement Learning with Lookahead Information (NeurIPS 2024 · similarity 0.88)
- Adaptive $Q$-Network: On-the-fly Target Selection for Deep Reinforcement Learning (ICLR 2025 · similarity 0.86)
- Horizon Adaptive Offline Policy Learning via Value Stitching (NeurIPS 2026 · similarity 0.85)
- Learning the Target Network in Function Space (ICML 2024 · similarity 0.85)
- Seizing Serendipity: Exploiting the Value of Past Success in Off-Policy Actor-Critic (ICML 2024 · similarity 0.85)
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.