Research map / Reinforcement learning / Imitation learning
Learning to Reach Goals via Iterated Supervised 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 GCSL instead of TD3-HER | Goal-reaching policies · Robotic pushing and claw manipulation | distance to goal | Helped |
| Use GCSL instead of the strongest prior RL method | Goal-reaching policies · Lunar Lander | distance to goal | Helped |
| Use GCSL instead of PPO | Goal-reaching policies · Door Opening | distance to goal | Hurt |
| Use GCSL instead of Hindsight Policy Gradients | Goal-reaching policies · Fetch-Reach | distance to goal | Helped |
| Condition the policy on the remaining horizon (ablation) | GCSL · Lunar Lander | training speed | Helped |
| Condition the policy on the remaining horizon (ablation) | GCSL · Sawyer pushing | goal-reaching performance | Hurt |
| Collect data with a fixed uniform policy instead of the latest learned policy (ablation) | GCSL · Goal-reaching tasks | learning speed | Hurt |
| Train only on recent on-policy data instead of reusing the full dataset (ablation) | GCSL · Goal-reaching tasks | learning speed | Hurt |
| Limit relabeling to goals at most three steps away (ablation) | GCSL · Goal-reaching tasks | goal-reaching performance | Hurt |
| Add expert demonstrations to the initial dataset | GCSL · Robotic pushing | learning speed | Helped |
| Use GCSL rather than TD3-HER to learn from expert demonstrations | Goal-reaching policies · Robotic pushing | performance beyond the behavior-cloned policy | Helped |
| Use GCSL rather than TD3-HER | Goal-reaching policies · Hyperparameter sweep across nine configurations | sensitivity to hyperparameters | Helped |
Closest papers by meaning
- Learning from Demonstrations via Capability-Aware Goal Sampling (NeurIPS 2025 · similarity 0.89)
- GO-DICE: Goal-Conditioned Option-Aware Offline Imitation Learning via Stationary Distribution Correction Estimation (AAAI 2024 · similarity 0.88)
- Learning Process Rewards via Visitation Matching for Efficient RL (NeurIPS 2026 · similarity 0.88)
- Stabilizing Contrastive RL: Techniques for Robotic Goal Reaching from Offline Data (ICLR 2024 · similarity 0.88)
- Enhancing Online Reinforcement Learning with Meta-Learned Objective from Offline Data (AAAI 2025 · similarity 0.88)
- Imitation Learning from Observation with Automatic Discount Scheduling (ICLR 2024 · 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.