Research map / Reinforcement learning / Visual and representation RL
Continuous Adaptation via Meta-Learning in Nonstationary and Competitive Environments
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 LSTM meta-updates instead of the strongest listed baseline | Ant · RoboSumo vs Ant | win rate: 30.0 -> 44.0 | Helped |
| Use LSTM meta-updates instead of the strongest listed baseline | Ant · RoboSumo vs Bug | win rate: 21.0 -> 34.6 | Helped |
| Use LSTM meta-updates instead of the strongest listed baseline | Ant · RoboSumo vs Spider | win rate: 27.6 -> 35.1 | Helped |
| Use LSTM meta-updates instead of the strongest listed baseline | Bug · RoboSumo vs Ant | win rate: 33.5 -> 39.5 | Helped |
| Use LSTM meta-updates instead of the strongest listed baseline | Bug · RoboSumo vs Bug | win rate: 28.6 -> 43.7 | Helped |
| Use LSTM meta-updates instead of the strongest listed baseline | Bug · RoboSumo vs Spider | win rate: 45.8 -> 52.0 | Helped |
| Use LSTM meta-updates instead of the strongest listed baseline | Spider · RoboSumo vs Ant | win rate: 48.0 -> 45.3 | Hurt |
| Use LSTM meta-updates instead of the strongest listed baseline | Spider · RoboSumo vs Bug | win rate: 43.9 -> 48.4 | Helped |
| Use LSTM meta-updates instead of the strongest listed baseline | Spider · RoboSumo vs Spider | win rate: 42.2 -> 46.7 | Helped |
| Use meta-learned adaptation rather than tracking | nonstationary locomotion · Held-out leg-torque changes | episodic reward | No clear effect |
| Use meta-learned adaptation rather than reactive baselines | continuous adaptation methods · Few-shot nonstationary and competitive environments | adaptation performance | Helped |
| Use meta-learned adaptation rather than tracking | RoboSumo agents · More than 50 episodes per round | win rate | Mixed |
Closest papers by meaning
- Learning to Adapt in Dynamic, Real-World Environments through Meta-Reinforcement Learning (ICLR 2019 · similarity 0.90)
- AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers (NeurIPS 2024 · similarity 0.88)
- Enabling Adaptive Agent Training in Open-Ended Simulators by Targeting Diversity (NeurIPS 2024 · similarity 0.88)
- Constrained Meta-Reinforcement Learning for Adaptable Safety Guarantee with Differentiable Convex Programming (AAAI 2024 · similarity 0.87)
- Meta-Reinforcement Learning with Universal Policy Adaptation: Provable Near-Optimality under All-task Optimum Comparator (NeurIPS 2024 · similarity 0.87)
- Skill-aware Mutual Information Optimisation for Zero-shot Generalisation in Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
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.