Research map / Reinforcement learning / Visual and representation RL
LESSON: Learning to Integrate Exploration Strategies for Reinforcement Learning via an Option Framework
Woojun Kim, Jeonghye Kim, Youngchul Sung
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 |
|---|---|---|---|
| Adaptively integrate greedy, random, temporally extended random, and intrinsic-reward exploration policies using an option framework | DQN-based reinforcement learning · MiniGrid and Atari 2600 environments | performance | Improved |
| Replace adaptive selection with equal-probability selection among exploration strategies | DQN exploration · MiniGrid and Atari 2600 environments | performance | Hurt |
| Learn options from scratch instead of using predefined intra-policies (ablation) | LESSON · MiniGrid | performance | Hurt |
| Remove the greedy policy from the intra-policy set (ablation) | LESSON · MiniGrid tasks | performance | Hurt |
| Remove the most effective exploration policy for an environment (ablation) | LESSON · MiniGrid tasks | performance | Hurt |
| Set the intrinsic-reward coefficient too low (ablation) | LESSON behavior policy · Sparse-reward tasks | learning speed | Hurt |
| Set the intrinsic-reward coefficient too high (ablation) | LESSON behavior policy · Tasks with extrinsic rewards | learning speed | Hurt |
| Replace the RND-based exploration policy with a PixelCNN pseudo-count policy | LESSON · Montezuma's Revenge | learning speed | Improved |
| Use a learned mixture of randomness-based and history-based exploration rather than PixelCNN count-based exploration alone | LESSON · Montezuma's Revenge | performance | Improved |
Closest papers by meaning
- Flexible Option Learning (NeurIPS 2021 · similarity 0.87)
- Reward-Free Exploration for Reinforcement Learning (ICML 2020 · similarity 0.87)
- OptionZero: Planning with Learned Options (ICLR 2025 · similarity 0.86)
- Exploration from a Primal-Dual Lens: Value-Incentivized Actor-Critic Methods for Sample-Efficient Online RL (NeurIPS 2025 · similarity 0.85)
- Autonomous Option Invention for Continual Hierarchical Reinforcement Learning and Planning (AAAI 2025 · similarity 0.85)
- Beyond Optimism: Exploration With Partially Observable Rewards (NeurIPS 2024 · 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.
