Research map / Reinforcement learning / Visual and representation RL
Leveraging Explanation to Improve Generalization of Meta Reinforcement Learning
Shicheng Liu, Minghui Zhu
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 XMRL critical-task augmentation | MAML · Drone navigation | success rate: 0.87 -> 0.97 | Improved |
| add XMRL critical-task augmentation | MAML · Stock market | cumulative reward: 359.13 -> 421.13 | Improved |
| add XMRL critical-task augmentation | MAML · HalfCheetah | cumulative reward: -68.89 -> -44.67 | Improved |
| add XMRL critical-task augmentation | MAML · Ant | cumulative reward: 100.64 -> 119.15 | Improved |
| add task weighting instead of augmentation | MAML · Drone navigation | success rate: 0.87 -> 0.87 | No clear effect |
| add meta augmentation with predefined mixing distribution instead of learned one | MAML · Stock market | cumulative reward: 359.13 -> 389.17 | Improved |
| add meta regularization instead of XMRL | MAML · Ant | cumulative reward: 100.64 -> 104.15 | Improved |
| add XMRL, performance on poorly adapted tasks (ablation) | MAML · Drone navigation, poorly adapted tasks | success rate: 0.55 -> 0.93 | Improved |
| add XMRL, performance on poorly adapted tasks (ablation) | MAML · Stock market, poorly adapted tasks | cumulative reward: 71.05 -> 381.33 | Improved |
| add XMRL, performance on poorly adapted tasks (ablation) | MAML · HalfCheetah, poorly adapted tasks | cumulative reward: -162.09 -> -55.00 | Improved |
| add XMRL, performance on non-poorly adapted tasks (ablation) | MAML · Stock market, non-poorly adapted tasks | cumulative reward: 431.15 -> 431.08 | No clear effect |
| add XMRL, performance on non-poorly adapted tasks (ablation) | MAML · Ant, non-poorly adapted tasks | cumulative reward: 115.88 -> 124.02 | Improved |
Closest papers by meaning
- Meta-Reinforcement Learning with Universal Policy Adaptation: Provable Near-Optimality under All-task Optimum Comparator (NeurIPS 2024 · similarity 0.91)
- AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers (NeurIPS 2024 · similarity 0.89)
- Can Learned Optimization Make Reinforcement Learning Less Difficult? (NeurIPS 2024 · similarity 0.89)
- Skill-aware Mutual Information Optimisation for Zero-shot Generalisation in Reinforcement Learning (NeurIPS 2024 · similarity 0.89)
- Meta-Reinforcement Learning with Adaptation from Human Feedback via Preference-Order-Preserving Task Embedding (ICML 2025 · similarity 0.89)
- MetaCARD: Meta-Reinforcement Learning with Task Uncertainty Feedback via Decoupled Context-Aware Reward and Dynamics Components (AAAI 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.
