Research map / Reinforcement learning / Offline RL
Goal-Conditioned Predictive Coding for 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 GCPC instead of IQL | Offline RL on AntMaze · AntMaze average | normalized score: 57.3 -> 69.4 | Helped |
| Use GCPC instead of WGCSL | Offline RL on Kitchen · Kitchen average | normalized score: 76.5 -> 82.9 | Helped |
| Use GCPC instead of Decision Diffuser | Offline RL on Gym Locomotion · Gym Locomotion average | normalized score: 81.8 -> 79.7 | Hurt |
| Use MAE-RC predictive coding instead of history-only MAE-H (ablation) | GCPC two-stage pretraining without actions · AntMaze Large-Play | normalized score: 36.6 -> 78.2 | Helped |
| Remove actions from trajectory pretraining (ablation) | GCPC with MAE-RC · AntMaze Large-Play | normalized score: 62.0 -> 78.2 | Helped |
| Add goal conditioning to TrajNet (ablation) | GCPC with MAE-RC · AntMaze Large-Play | normalized score: 16.0 -> 78.2 | Helped |
| Use latent future bottleneck instead of decoded explicit future states (ablation) | GCPC policy conditioning · AntMaze Large-Play | normalized score: 67.9 -> 78.2 | Helped |
| Use a longer future window, p=70 instead of p=20 (ablation) | GCPC · AntMaze Large-Diverse | normalized score: 62.4 -> 80.6 | Helped |
| Use 4 slot tokens instead of 2 (ablation) | GCPC · AntMaze Large-Diverse | normalized score: 77.2 -> 80.6 | Helped |
| Use MAE-F rather than MAE-RC (ablation) | GCPC two-stage pretraining without actions · AntMaze Large-Play | normalized score: 78.2 -> 76.2 | Hurt |
| Use MAE-F rather than MAE-RC (ablation) | GCPC two-stage pretraining without actions · Kitchen-Partial | normalized score: 90.2 -> 92.5 | Helped |
| Include actions in trajectory pretraining (ablation) | GCPC with MAE-F · Kitchen-Partial | normalized score: 86.4 -> 92.5 | Hurt |
Closest papers by meaning
- Stitching Sub-trajectories with Conditional Diffusion Model for Goal-Conditioned Offline RL (AAAI 2024 · similarity 0.90)
- Are Expressive Models Truly Necessary for Offline RL? (AAAI 2025 · similarity 0.89)
- Foundation Policies with Hilbert Representations (ICML 2024 · similarity 0.89)
- Flattening Hierarchies with Policy Bootstrapping (NeurIPS 2025 · similarity 0.89)
- Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models (NeurIPS 2025 · similarity 0.88)
- Offline Goal-conditioned Reinforcement Learning with Quasimetric Representations (NeurIPS 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.