Research map / Reinforcement learning / Imitation learning
Unlabeled Imperfect Demonstrations in Adversarial Imitation 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 |
|---|---|---|---|
| Apply UID to WAIL | WAIL · MuJoCo D1 and D2: Ant-v2, HalfCheetah-v2, Walker2d-v2 | average cumulative reward: Ant-v2 D1: 1348$\pm$120 -> 1709$\pm$118; HalfCheetah-v2 D1: 2282$\pm$58 -> 2569$\pm$157; Walker2d-v2 D1: 2180$\pm$46 -> | Helped |
| Apply UID to GAIL | GAIL · MuJoCo D1 and D2: Ant-v2, HalfCheetah-v2, Walker2d-v2 | average cumulative reward: Ant-v2 D1: 1179$\pm$158 -> 1674$\pm$142; HalfCheetah-v2 D1: 2159$\pm$139 -> 3276$\pm$114; Walker2d-v2 D1: 1873$\pm$115 - | Helped |
| Use UID instead of BCND as the optimal-demonstration ratio decreases below 25% | Ant-v2 · Ant-v2, varying ratios of optimal demonstrations | performance | Helped |
| Set α to 0.7 rather than other tested values (ablation) | UID · Ant-v2, varying ratios of optimal demonstrations | performance | Helped |
| Vary α across 0.3, 0.5, 0.7, and 0.5-0.7 (ablation) | UID · Ant-v2, varying ratios of optimal demonstrations | performance | No clear effect |
| Treat expert demonstrations as unlabeled and use PU discrimination to dynamically match agent trajectories | UID-GAIL versus GAIL · MuJoCo D1 and D2: Ant-v2, HalfCheetah-v2, Walker2d-v2 | discriminator accuracy on optimal and non-optimal demonstrations | Helped |
Closest papers by meaning
- PN-GAIL: Leveraging Non-optimal Information from Imperfect Demonstrations (ICLR 2025 · similarity 0.92)
- Discriminator-Actor-Critic: Addressing Sample Inefficiency and Reward Bias in Adversarial Imitation Learning (ICLR 2019 · similarity 0.91)
- DualCOIL: Offline Imitation Learning from Contrasting Demonstrations (ICML 2026 · similarity 0.91)
- Learning to Weight Imperfect Demonstrations (ICML 2021 · similarity 0.91)
- Imitation Learning from Imperfect Demonstration (ICML 2019 · similarity 0.91)
- Adversarial Imitation Learning via Boosting (ICLR 2024 · similarity 0.89)
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.