Research map / Reinforcement learning / Visual and representation RL
Accelerating Exploration with Unlabeled Prior Data
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 |
|---|---|---|---|
| Label prior transitions with optimistic UCB rewards instead of unbiased reward estimates | ExPLORe with RLPD · AntMaze | normalized return | Helped |
| Label prior transitions with optimistic UCB rewards instead of unbiased reward estimates | ExPLORe with RLPD · COG | normalized return | Helped |
| Add an ICVF-pretrained encoder initialization to the reward and RND models (ablation) | ExPLORe · COG image-input tasks | online learning efficiency | Helped |
| Initialize with ICVF representations but omit optimistic reward labeling (ablation) | Naïve reward labeling · COG image-input tasks | online learning success | No clear effect |
| Add an RND bonus to online data in addition to optimistic labeling of prior data (ablation) | ExPLORe · AntMaze large-diverse-v2, insufficient-coverage corruption | success | Helped |
| Add an RND bonus to online data in addition to optimistic labeling of prior data (ablation) | ExPLORe · AntMaze large-diverse-v2, orthogonal-transitions corruption | online learning performance | Helped |
| Periodically reset the reward model and RND networks (ablation) | ExPLORe · Adroit relocate | normalized return | Helped |
| Remove the RND bonus from optimistic reward labels (ablation) | ExPLORe with IQL · AntMaze medium datasets | success | Hurt |
| Use an unbiased learned reward estimate rather than optimistic UCB rewards (ablation) | ExPLORe · AntMaze state coverage | state coverage | Hurt |
| Use behavior cloning to follow prior-data actions alongside naïve reward labeling (ablation) | Naïve Reward Labeling · COG | online learning success | No clear effect |
| Use optimistic reward labeling on prior data rather than novelty bonuses only on online data | Online RL with RND · AntMaze and COG | online learning performance | Helped |
Closest papers by meaning
- Leveraging Skills from Unlabeled Prior Data for Efficient Online Exploration (ICML 2025 · similarity 0.92)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.90)
- Reinforcement Learning with Action Chunking (NeurIPS 2025 · similarity 0.88)
- Beyond Optimism: Exploration With Partially Observable Rewards (NeurIPS 2024 · similarity 0.88)
- Decoupling Exploration and Policy Optimization: Uncertainty Guided Tree Search for Hard Exploration (NeurIPS 2026 · similarity 0.88)
- MaxInfoRL: Boosting exploration in reinforcement learning through information gain maximization (ICLR 2025 · 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.