Research map / Reinforcement learning / Visual and representation RL
Diversity is All You Need: Learning Skills without a Reward Function
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 maximum-entropy skills rather than lower-entropy skills (ablation) | DIAYN imitation learning · 600 imitation tasks | trajectory distance | Helped |
| Learn 50 skills rather than only 5 (ablation) | DIAYN imitation learning · 600 imitation tasks | trajectory distance | Helped |
| Fix the skill prior instead of learning it (ablation) | DIAYN · HalfCheetah, InvertedPendulum, MountainCar | effective number of skills: 10x | Helped |
| Increase entropy regularization (ablation) | DIAYN · 2D point mass | state-space diversity | Mixed |
| Use DIAYN skill pretraining to initialize policy and value networks (ablation) | SAC task fine-tuning · HalfCheetah, Hopper, Ant | learning speed | Helped |
| Condition the discriminator on prior-selected observation features (ablation) | DIAYN hierarchical RL · Ant navigation | performance | Helped |
| Compose DIAYN skills with a learned meta-controller | Hierarchical RL · Cheetah hurdle and Ant navigation | task reward | Helped |
| Use a collection of DIAYN policies rather than one exploration-bonus policy | Exploration · Running, jumping, and moving away from origin | task reward | Helped |
Closest papers by meaning
- Exploration by Learning Diverse Skills through Successor State Representations (NeurIPS 2024 · similarity 0.91)
- Skill Disentanglement in Reproducing Kernel Hilbert Space (AAAI 2025 · similarity 0.90)
- Acquiring Diverse Skills using Curriculum Reinforcement Learning with Mixture of Experts (ICML 2024 · similarity 0.90)
- Efficient Skill Discovery via Regret-Aware Optimization (ICML 2025 · similarity 0.89)
- Periodic Skill Discovery (NeurIPS 2025 · similarity 0.89)
- Constrained Ensemble Exploration for Unsupervised Skill Discovery (ICML 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.