Research map / Reinforcement learning / Visual and representation RL
Human-Timescale Adaptation in an Open-Ended Task Space
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 AdA for multi-trial adaptation instead of a single-trial-trained baseline | AdA · XLand single-agent held-out test tasks | last-trial score | Helped |
| Use Transformer-XL memory instead of a vanilla recurrent network (ablation) | AdA · XLand single-agent adaptation tasks | last-trial score | Helped |
| Add attention memory to a recurrent network (ablation) | AdA · XLand single-agent adaptation tasks | last-trial score | Helped |
| Use no-op filtering instead of uniform task sampling (ablation) | AdA · XLand single-agent test tasks | last-trial score | Helped |
| Use prioritized level replay instead of uniform task sampling (ablation) | AdA · XLand single-agent test tasks | last-trial score | Helped |
| Use prioritized level replay instead of no-op filtering (ablation) | AdA · XLand single-agent test tasks, especially at higher trial counts | last-trial score | Helped |
| Increase Transformer parameter count (ablation) | AdA · XLand held-out test tasks | test score | Helped |
| Increase Transformer-XL memory length (ablation) | AdA · XLand held-out test tasks | test score | Helped |
| Increase the size of the training task pool (ablation) | AdA · XLand held-out test tasks | test score | Helped |
| Increase the complexity of the training task distribution (ablation) | AdA · XLand test tasks | test score | Helped |
| Kickstart training with teacher distillation (ablation) | AdA · XLand multi-agent test tasks | test score | Helped |
| Train with a larger maximum number of trials (ablation) | AdA · XLand held-out test tasks at long adaptation horizons | last-trial score | Helped |
Closest papers by meaning
- Enabling Adaptive Agent Training in Open-Ended Simulators by Targeting Diversity (NeurIPS 2024 · similarity 0.90)
- Behavioral Exploration: Learning to Explore via In-Context Adaptation (ICML 2025 · similarity 0.89)
- SEEA-R1: Tree-Structured Reinforcement Fine-Tuning for Self-Evolving Embodied Agents (NeurIPS 2025 · similarity 0.88)
- SimWorld Studio: Automatic Environment Generation with Evolving Coding Agent for Embodied Agent Learning (NeurIPS 2026 · similarity 0.88)
- AMAGO-2: Breaking the Multi-Task Barrier in Meta-Reinforcement Learning with Transformers (NeurIPS 2024 · similarity 0.88)
- Kinetix: Investigating the Training of General Agents through Open-Ended Physics-Based Control Tasks (ICLR 2025 · 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.