Research map / Reinforcement learning / Visual and representation RL
Adaptive Skills Adaptive Partitions (ASAP)
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 |
|---|---|---|---|
| Learn skills and skill partitions jointly from an initially misspecified model | ASAP · RoboCup R1 | goal-scoring ratio: 18% -> 79% | Helped |
| Use learned ASAP policy instead of the approximately optimal scoring controller | RoboCup R1 · RoboCup R1 | goal-scoring ratio: 91% -> 79% | Hurt |
| Increase the number of skill hyperplanes (ablation) | ASAP · Two Rooms | average reward | No clear effect |
| Initialize the new task with skills and partitions learned on the first task | ASAP · Flipped Two Rooms (Task 2) | episodes to near-optimal performance: 35000 -> 5000 | Helped |
| Flip the signs of learned hyperplane parameters without additional learning (ablation) | ASAP · Flipped Two Rooms | average reward | Helped |
| Use a linear rather than degree-two polynomial hyperplane feature representation (ablation) | ASAP · RoboCup R1 | average reward | Helped |
| Add the defender's x-location as a hyperplane feature (ablation) | ASAP · RoboCup R2 | average reward | Mixed |
| Add the defender's y-location as a hyperplane feature (ablation) | ASAP · RoboCup R2 | average reward | Mixed |
| Add the defender's x- and y-location as hyperplane features (ablation) | ASAP · RoboCup R2 | average reward | Mixed |
| Reuse a learned skill in two distinct state-space regions (ablation) | ASAP · Three Rooms | average reward | Helped |
| Learn skills and partitions rather than execute the fixed initial misspecified partitioning (ablation) | ASAP · Two Rooms | average reward | Helped |
| Use more skills by increasing the hyperplane count (ablation) | ASAP · Two Rooms | computational complexity | Hurt |
Closest papers by meaning
- Skill Expansion and Composition in Parameter Space (ICLR 2025 · similarity 0.87)
- Skill-Pro: Learning Reusable Skills from Experience via Non-Parametric PPO for LLM Agents (ICML 2026 · similarity 0.87)
- SkillOS: Learning Skill Curation for Self-Evolving Agents (NeurIPS 2026 · similarity 0.86)
- Think Small, Act Big: Primitive Prompt Learning for Lifelong Robot Manipulation (CVPR 2025 · similarity 0.86)
- Policy Compatible Skill Incremental Learning via Lazy Learning Interface (NeurIPS 2025 · similarity 0.85)
- Learning to Continually Learn via Meta-learning Agentic Memory Designs (NeurIPS 2026 · 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.