Research map / Reinforcement learning / Imitation learning
Adversarial Imitation Learning With General Function Approximation: Theoretical Analysis and Practical Algorithms
Tian Xu, Zhilong Zhang, Zexuan Chen, Ruishuo Chen, Yihao Sun, Yang Yu
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 |
|---|---|---|---|
| add optimism regularization to policy learning (ablation) | MF OPT-AIL and MB OPT-AIL · DMControl 8 tasks, 1 expert demo | learning curve convergence | Improved |
| use MF OPT-AIL instead of HyPE | HyPE · DMControl average, 1 demo | policy return: 534.0 -> 683.0 | Improved |
| use MB OPT-AIL instead of HyPER | HyPER · DMControl average, 1 demo | policy return: 613.4 -> 691.4 | Improved |
| use MF OPT-AIL instead of BC | BC · DMControl average, 1 demo | policy return: 82.0 -> 683.0 | Improved |
| increase expert demos from 1 to 10 | MF OPT-AIL · DMControl average | policy return: 683.0 -> 755.6 | Improved |
| increase expert demos from 1 to 10 | MB OPT-AIL · DMControl average | policy return: 691.4 -> 700.0 | No clear effect |
| increase expert demos from 1 to 10 | CMIL · DMControl average | policy return: 389.7 -> 450.2 | Mixed |
| use OPT-AIL reward update instead of FILTER | FILTER · DMControl average | Pearson correlation: 0.5868 -> 0.7763 | Improved |
| use MB OPT-AIL reward update instead of CMIL | CMIL · DMControl average | Pearson correlation: 0.5748 -> 0.7472 | Improved |
| use model-based rather than model-free learning | OPT-AIL · DMControl 1 expert demo | interaction efficiency | Improved |
| use MB OPT-AIL on Cheetah Run instead of MF OPT-AIL | MF OPT-AIL · DMControl Cheetah Run, 1 demo | policy return: 348.7 -> 577.0 | Improved |
| use MB OPT-AIL instead of MF OPT-AIL | MF OPT-AIL · DMControl average, 10 demos | policy return: 755.6 -> 700.0 | Hurt |
Closest papers by meaning
- Provably and Practically Efficient Adversarial Imitation Learning with General Function Approximation (NeurIPS 2024 · similarity 0.98)
- Provably Efficient Policy-Reward Co-Pretraining for Adversarial Imitation Learning (ICML 2026 · similarity 0.93)
- On Discovering Algorithms for Adversarial Imitation Learning (ICLR 2026 · similarity 0.92)
- Near-Optimal Second-Order Guarantees for Model-Based Adversarial Imitation Learning (ICLR 2026 · similarity 0.92)
- Adversarial Imitation Learning via Boosting (ICLR 2024 · similarity 0.92)
- Exploring Gradient Explosion in Generative Adversarial Imitation Learning: A Probabilistic Perspective (AAAI 2024 · similarity 0.90)
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.
