Research map / Reinforcement learning / Imitation learning
Policy Optimization with Demonstrations
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 |
|---|---|---|---|
| POfD vs TRPO expert trained in dense reward | POfD · Walker2d-v1 | empirical return: 6717.08 -> 7687.47 | Helped |
| POfD vs expert on HalfCheetah | POfD · HalfCheetah-v1 | empirical return: 4463.46 -> 4771.15 | Helped |
| POfD vs expert on Hopper | POfD · Hopper-v1 | empirical return: 3571.38 -> 3652.23 | Helped |
| POfD vs expert on Reacher (naturally sparse, 15 demos) | POfD · Reacher-v1 | empirical return: 0.75 -> 0.86 | Helped |
| POfD vs expert on Humanoid (376-dim state) | POfD · Humanoid-v1 | empirical return: 9575.40 -> 9823.43 | Helped |
| POfD vs expert on CartPole | POfD · CartPole-v0 | empirical return: 500 -> 500 | No clear effect |
| POfD vs expert on MountainCar | POfD · MountainCar-v1 | empirical return: −98.75 -> −98.35 ± 9.35 | No clear effect |
| POfD exceeds the demonstration quality | POfD · Walker2d-v1 | empirical return: 1701.13 -> 7687.47 | Helped |
| Add demonstration-guided occupancy matching to TRPO under sparse reward | TRPO · Sparse locomotion tasks | learning curves | Helped |
| POfD vs GAIL under same demonstration | POfD · MountainCar | learning curves | Helped |
| POfD vs DDPGfD and GAIL on Humanoid | POfD · Humanoid-v1 | learning curves | Helped |
| DQfD with a single demonstration on discrete tasks | DQfD · CartPole, MountainCar | learning curves | Hurt |
Closest papers by meaning
- Uncertainty-Based Smooth Policy Regularisation for Reinforcement Learning with Few Demonstrations (NeurIPS 2025 · similarity 0.88)
- DISCOVER: Automated Curricula for Sparse-Reward Reinforcement Learning (NeurIPS 2025 · similarity 0.87)
- Action-Dependent Optimality-Preserving Reward Shaping (ICML 2025 · similarity 0.87)
- Learning Process Rewards via Visitation Matching for Efficient RL (NeurIPS 2026 · similarity 0.87)
- Beyond-Expert Performance with Limited Demonstrations: Efficient Imitation Learning with Double Exploration (ICLR 2025 · similarity 0.87)
- A Geometric Perspective on Reward Function Updates in Inverse Reinforcement Learning (NeurIPS 2026 · 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.