Research map / Reinforcement learning / Visual and representation RL
PlanGAN: Model-based Planning With Sparse Rewards and Multiple Goals
Henry Charlesworth, Giovanni Montana
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (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 the planner instead of directly using GAN-proposed actions (NoPlanner, NoPlannerAvg) (ablation) | PlanGAN · Fetch Push and Fetch Pick And Place | success rate | Improved |
| use an ensemble of GANs (3) instead of a single GAN (ablation) | PlanGAN · Fetch Push and Fetch Pick And Place | success rate | Improved |
| increase ensemble from 3 to 5 GANs (ablation) | PlanGAN · Fetch Push and Fetch Pick And Place | success rate | Improved |
| add one-step model regularisation (lambda 30 vs 0) (ablation) | PlanGAN · Fetch Push and Fetch Pick And Place | success rate | Improved |
| use PlanGAN instead of HER | goal-conditioned sparse-reward control · Four Rooms, Reacher, Fetch Push, Fetch Pick And Place | sample efficiency: 4-8 times | Improved |
Closest papers by meaning
- Diffusion Subgoal Planning for Long-Horizon Offline Goal-Conditioned Reinforcement Learning (NeurIPS 2026 · similarity 0.88)
- Learning from Reward-Free Offline Data: A Case for Planning with Latent Dynamics Models (NeurIPS 2025 · similarity 0.87)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Learning World Models for Unconstrained Goal Navigation (NeurIPS 2024 · similarity 0.87)
- GenPlan: Generative Sequence Models as Adaptive Planners (AAAI 2025 · similarity 0.87)
- Flattening Hierarchies with Policy Bootstrapping (NeurIPS 2025 · similarity 0.87)
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.
