Research map / Reinforcement learning / Visual and representation RL
Temporally-Extended ε-Greedy Exploration
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 |
|---|---|---|---|
| Add εz-greedy exploration | R2D2 · Atari-57 | Median human-normalized score: 16.44 -> 17.77 | Helped |
| Add εz-greedy exploration | R2D2 · Atari-57 | Human-gap: 0.102 -> 0.077 | Helped |
| Add εz-greedy exploration | Rainbow · Atari-57 | Median human-normalized score: 2.03 -> 2.30 | Helped |
| Add εz-greedy exploration | Rainbow · Atari-57 | Mean human-normalized score: 8.82 -> 9.34 | Helped |
| Use RND intrinsic-reward exploration | R2D2 · Atari-57 | Median human-normalized score: 16.44 -> 11.14 | Hurt |
| Use RND intrinsic-reward exploration | R2D2 · Atari-57 | Mean human-normalized score: 39.55 -> 42.17 | Helped |
| Use RND intrinsic-reward exploration | R2D2 · Atari-57 | Human-gap: 0.102 -> 0.151 | Hurt |
| Use Bootstrapped DQN exploration | R2D2 · Atari-57 | Median human-normalized score: 16.44 -> 16.62 | Helped |
| Add εz-greedy to Rainbow | Rainbow · Atari-57 | Human-gap: 0.139 -> 0.130 | Helped |
| Replace the zeta duration distribution with exponential or uniform durations (ablation) | εz-greedy · Chain MDP and six Atari games | Average return | Mixed |
| Use zeta-distributed action-repeat durations rather than one-step ε-greedy actions | ε-greedy · DeepSea, GridWorld, MountainCar, and CartPole swingup-sparse | Average return | Helped |
| Use zeta-duration action repeats when action effects are randomized by state (ablation) | εz-greedy · DeepSea with randomized action effects | Exploration efficiency | No clear effect |
Closest papers by meaning
- Emergence of Exploration in Policy Gradient Reinforcement Learning via Retrying (ICML 2026 · similarity 0.90)
- Explicit Explore-Exploit Algorithms in Continuous State Spaces (NeurIPS 2019 · similarity 0.90)
- Reward-Free Exploration for Reinforcement Learning (ICML 2020 · similarity 0.90)
- Beyond Optimism: Exploration With Partially Observable Rewards (NeurIPS 2024 · similarity 0.90)
- Fast active learning for pure exploration in reinforcement learning (ICML 2021 · similarity 0.89)
- Decoupling Exploration and Policy Optimization: Uncertainty Guided Tree Search for Hard Exploration (NeurIPS 2026 · 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.