Research map / Reinforcement learning / Visual and representation RL
Exploration via Hindsight 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 |
|---|---|---|---|
| Generate hindsight goals by matching target tasks to achieved states that are both reachable and close to the target distribution | DDPG+HER · Eight robotic manipulation environments | sample efficiency: four ... four | Mixed |
| Add Energy-Based Prioritization to HGG | HGG · Six environments eligible for EBP | performance: six environments | Helped |
| Use a crafted obstacle-aware distance metric instead of the ℓ2 distance | HGG · FetchPush with obstacle | sample efficiency | Helped |
| Generate hindsight goals with HGG instead of HER | HER and HER+GOID · Fixed-initial-state curriculum comparison environment | performance | Helped |
| Use HGG with different Lipschitz constant L values, provided L is not too small (ablation) | HGG · FetchPush | performance | No clear effect |
| Vary the distance weight c (ablation) | HGG · Robotic manipulation tasks tested | performance | No clear effect |
| Vary the number of hindsight goals K (ablation) | HGG · Simple tasks | performance | No clear effect |
| Use a larger batch size K (ablation) | HGG · FetchPickAndPlace | variance of training results | Helped |
| Limit each trajectory to at most one selected hindsight state (ablation) | HGG · Robotic manipulation tasks | robustness | Helped |
Closest papers by meaning
- DHER: Hindsight Experience Replay for Dynamic Goals (ICLR 2019 · similarity 0.91)
- Hindsight Experience Replay (NeurIPS 2017 · similarity 0.91)
- Cluster-Based Sampling in Hindsight Experience Replay for Robotic Tasks (Student Abstract) (AAAI 2024 · similarity 0.88)
- Goal-Conditioned On-Policy Reinforcement Learning (NeurIPS 2024 · similarity 0.87)
- A Single Goal is All You Need: Skills and Exploration Emerge from Contrastive RL without Rewards, Demonstrations, or Subgoals (ICLR 2025 · similarity 0.84)
- How Does Goal Relabeling Improve Sample Efficiency? (ICML 2024 · similarity 0.84)
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.