Research map / Reinforcement learning / Visual and representation RL
Outcome-directed Reinforcement Learning by Uncertainty \& Temporal Distance-Aware Curriculum Goal Generation
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 OUTPACE instead of prior curriculum methods | Outcome-directed RL · U-Maze, N-Maze, Spiral-Maze, Ant Locomotion, Sawyer Push, Sawyer Pick&Place | distance from curriculum goals to final goals | Helped |
| Use OUTPACE instead of prior curriculum methods | Outcome-directed RL · U-Maze, N-Maze, Spiral-Maze, Ant Locomotion, Sawyer Push, Sawyer Pick&Place | episode success rate | Helped |
| Add both uncertainty-based and temporal-distance-based curriculum objectives (ablation) | OUTPACE · Ant Locomotion and Sawyer Pick&Place | distance from curriculum goals to desired final goals | Helped |
| Add temporal-distance objective to uncertainty-only curriculum (only-cnml) (ablation) | OUTPACE · Ant Locomotion and Sawyer Pick&Place | distance from curriculum goals to desired final goals | Helped |
| Add uncertainty objective to temporal-distance-only curriculum (only-f) (ablation) | OUTPACE · Ant Locomotion and Sawyer Pick&Place | distance from curriculum goals to desired final goals | Helped |
| Replace sparse reward with timestep-proportionally shaped intrinsic reward (ablation) | OUTPACE · Ant Locomotion and Sawyer Pick&Place | episode success rate | Helped |
| Replace bipartite matching with a GAN-based goal generator (ablation) | OUTPACE · Ant Locomotion and Sawyer Pick&Place | distance from curriculum goals to desired final goals | Hurt |
| Reduce guidance coefficient c below its default value (ablation) | OUTPACE · Ant Locomotion and Sawyer Pick&Place | distance from curriculum goals to desired final goals | Hurt |
| Add the guidance term η_guidance to uncertainty- and temporal-distance-based curriculum generation (ablation) | OUTPACE · Ant Locomotion and Sawyer Pick&Place | distance from curriculum goals to desired final goals | Helped |
Closest papers by meaning
- CQM: Curriculum Reinforcement Learning with a Quantized World Model (NeurIPS 2023 · similarity 0.90)
- Diversify \& Conquer: Outcome-directed Curriculum RL via Out-of-Distribution Disagreement (NeurIPS 2023 · similarity 0.89)
- Automatic Curriculum Learning through Value Disagreement (NeurIPS 2020 · similarity 0.88)
- Diffusion-based Curriculum Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
- Goal-Conditioned On-Policy Reinforcement Learning (NeurIPS 2024 · similarity 0.86)
- Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning (NeurIPS 2026 · similarity 0.86)
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.