Research map / Reinforcement learning / Visual and representation RL
Unsupervised Skill Discovery for Learning Shared Structures across Changing Environments
Sang-Hyun Lee, Seung-Woo Seo
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 |
|---|---|---|---|
| Incremental skill discovery with state-coverage-expanding objective and skill evaluation (full method) vs. prior unsupervised skill discovery baselines | skills for changing environments · maze2d-medium-v1 downstream | average return (100 episodes): 286.15 vs 127.67 (DISk), 22.63 (DADS), 7.96 (DIAYN) | Improved |
| Incremental skill discovery with state-coverage-expanding objective and skill evaluation (full method) vs. prior baselines | skills for changing environments · HalfCheetah-hurdle-tunnel-sequence-v3 downstream | average return (100 episodes): 608.69 vs 88.99 (DISk), 70.53 (DADS), 69.03 (DIAYN) | Improved |
| Incremental skill discovery with state-coverage-expanding objective and skill evaluation (full method) vs. prior baselines | skills for changing environments · HalfCheetah-tunnel-hurdle-sequence-v3 downstream | average return (100 episodes): 619.38 vs 304.47 (DISk), 237.90 (DADS), 63.83 (DIAYN) | Improved |
| Retaining previously learned skills (DISk) vs. not retaining (DIAYN/DADS) on maze downstream | DISk baseline · maze2d-medium-v1 downstream | average return (100 episodes): 127.67 vs 22.63 (DADS), 7.96 (DIAYN) | Improved |
| Skill evaluation process (similarity and entropy criteria) added to skill discovery (ablation) | skill set in HalfCheetah-continual-v3 · HalfCheetah-continual-v3 skill composition | state coverage (qualitative) | No clear effect |
| Skill evaluation process added to skill discovery in Ant-v3 stationary setting (ablation) | skill set in Ant-v3 · Ant-v3 skill diversity (qualitative) | behavior diversity (qualitative) | No clear effect |
Closest papers by meaning
- Disentangled Unsupervised Skill Discovery for Efficient Hierarchical Reinforcement Learning (NeurIPS 2024 · similarity 0.90)
- Skill Disentanglement in Reproducing Kernel Hilbert Space (AAAI 2025 · similarity 0.89)
- SkiLD: Unsupervised Skill Discovery Guided by Factor Interactions (NeurIPS 2024 · similarity 0.89)
- Periodic Skill Discovery (NeurIPS 2025 · similarity 0.88)
- Exploration by Learning Diverse Skills through Successor State Representations (NeurIPS 2024 · similarity 0.88)
- Task Adaptation from Skills: Information Geometry, Disentanglement, and New Objectives for Unsupervised Reinforcement Learning (ICLR 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.
