Research map / Reinforcement learning / Offline RL
When Data Geometry Meets Deep Function: Generalizing Offline Reinforcement Learning
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 DOGE's state-conditioned distance constraint instead of the baseline method | Offline RL on D4RL locomotion tasks · Locomotion total | average normalized score: 726.1±31.7 -> 768.6±55.4 | Helped |
| Use DOGE's state-conditioned distance constraint instead of the baseline method | Offline RL on D4RL AntMaze tasks · AntMaze total | average normalized score: 389.0±26.5 -> 403.3±40.9 | Helped |
| Allow policy learning in generalizable OOD regions using a distance-based constraint | DOGE · AntMaze with critical-path data removed | performance | Helped |
| Use a more restrictive distance threshold (30% quantile) (ablation) | DOGE · D4RL tasks in the threshold ablation | performance | Hurt |
| Use a more tolerant distance threshold (ablation) | DOGE · D4RL tasks in the threshold ablation | performance | Helped |
| Vary the number of noise actions used to train the distance function (ablation) | DOGE · D4RL tasks in the N ablation | performance | No clear effect |
| Vary alpha by adding or subtracting 2.5 from its original value (ablation) | DOGE · D4RL tasks in the alpha ablation | performance | No clear effect |
| Vary the distance threshold G across the tested quantiles (ablation) | DOGE · D4RL tasks in the G ablation | performance | Mixed |
Closest papers by meaning
- Offline RL with Smooth OOD Generalization in Convex Hull and its Neighborhood (ICLR 2025 · similarity 0.89)
- On the Role of General Function Approximation in Offline Reinforcement Learning (ICLR 2024 · similarity 0.89)
- Neural Network Approximation for Pessimistic Offline Reinforcement Learning (AAAI 2024 · similarity 0.89)
- Doubly Mild Generalization for Offline Reinforcement Learning (NeurIPS 2024 · similarity 0.88)
- Horizon Reduction Makes RL Scalable (NeurIPS 2025 · similarity 0.88)
- On the Complexity of Offline Reinforcement Learning with Q*-Approximation and Partial Coverage (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.