Research map / Reinforcement learning / Imitation learning
On the Value of Myopic Behavior in Policy Reuse
Chenjia Bai, Kang Xu, Shuang Qiu, Haoran He, Bin Zhao, Zhen Wang, Wei Li, Xuelong Li
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 decoupled (disentangled) value evaluation of prior policies instead of a single task-policy value function for policy selection (ablation) | SMEC · MetaWorld and AntMaze (aggregated) | success rate | Improved |
| add truncated short-horizon evaluation (lower discount factor) for prior policies instead of the same horizon as the task policy (ablation) | SMEC · MetaWorld and AntMaze (aggregated) | success rate | Improved |
| add the UCB term to value-guided policy selection (ablation) | SMEC · MetaWorld and AntMaze (aggregated) | success rate | Improved |
| vary the UCB coefficient c (ablation) | SMEC · MetaWorld and AntMaze (aggregated) | success rate | No clear effect |
| shorten behavior length h to H/50 (ablation) | SMEC · MetaWorld and AntMaze (aggregated) | success rate | Hurt |
| enlarge the prior policy set from Set 1 to Set 4 (add more policies and task-specific policies) (ablation) | SMEC · MetaWorld StickPull | success rate | Improved |
| use SMEC instead of Scratch SAC with prior policies | SAC · MetaWorld (12 tasks) and AntMaze (3 tasks) | success rate | Improved |
| use advantage-based reuse (CUP, MAMBA) instead of SMEC | SAC · MetaWorld and AntMaze | success rate | Hurt |
| use aggregation-based MultiPolar instead of SMEC | SAC · MetaWorld and AntMaze | success rate | Hurt |
| use hierarchical SkillS instead of SMEC | SAC · MetaWorld vs AntMaze | success rate | Mixed |
Closest papers by meaning
- Sample Efficient Myopic Exploration Through Multitask Reinforcement Learning with Diverse Tasks (ICLR 2024 · similarity 0.89)
- MONA: Myopic Optimization with Non-myopic Approval Can Mitigate Multi-step Reward Hacking (ICML 2025 · similarity 0.86)
- Constructing an Optimal Behavior Basis for the Option Keyboard (NeurIPS 2025 · similarity 0.86)
- Reflective Prompted Policy Optimization: Trajectory-Grounded Revision and Salience Bias (NeurIPS 2026 · similarity 0.85)
- DUO: Diverse, Uncertain, On-Policy Query Generation and Selection for Reinforcement Learning from Human Feedback (AAAI 2025 · similarity 0.85)
- STAIR: Addressing Stage Misalignment through Temporal-Aligned Preference Reinforcement Learning (NeurIPS 2025 · 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.
