Research map / Reinforcement learning / Imitation learning
Off-Policy Imitation Learning from Observations
Zhuangdi Zhu, Kaixiang Lin, Bo Dai, Jiayu Zhou
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 |
|---|---|---|---|
| remove inverse action regularization (OPOLO-x) (ablation) | OPOLO · HalfCheetah | return: 7336.96 -> 7632.80 | Mixed |
| remove inverse action regularization (OPOLO-x) (ablation) | OPOLO · Ant | return: 5783.57 -> 5112.04 | Mixed |
| remove inverse action regularization (OPOLO-x) (ablation) | OPOLO · Hopper | return: 3517.39 -> 3581.85 | No clear effect |
| remove inverse action regularization (OPOLO-x) (ablation) | OPOLO · Walker | return: 3803.00 -> 3947.72 | No clear effect |
| remove inverse action regularization (OPOLO-x) (ablation) | OPOLO · Swimmer | return: 257.38 -> 246.62 | No clear effect |
| remove inverse action regularization (OPOLO-x) (ablation) | OPOLO · HalfCheetah and Ant | sample efficiency | Improved |
| use inverse action model as regularizer instead of behavior cloning (OPOLO vs BCO) | OPOLO · Ant | return: 1529.54 -> 5783.57 | Improved |
| use inverse action model as regularizer instead of behavior cloning (OPOLO vs BCO) | OPOLO · HalfCheetah | return: 3881.10 -> 7336.96 | Improved |
| learn from observations only (OPOLO vs ValueDICE-fO) | OPOLO · Walker | return: 431.00 -> 3803.00 | Improved |
| learn from observations only (OPOLO vs DAC-fO) | OPOLO · Walker | return: 3033.02 -> 3803.00 | Improved |
| no expert actions (OPOLO vs DAC with actions) | OPOLO · Ant | return: 5424.28 -> 5783.57 | No clear effect |
| vary f-function (p,q) in f-divergence (ablation) | OPOLO · Ant | return curve | No clear effect |
Closest papers by meaning
- Imitation from Observations with Trajectory-Level Generative Embeddings (NeurIPS 2026 · similarity 0.90)
- Diffusion Imitation from Observation (NeurIPS 2024 · similarity 0.89)
- Imitation Learning from Observation with Automatic Discount Scheduling (ICLR 2024 · similarity 0.89)
- Learning on One Mode: Addressing Multi-modality in Offline Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Limited Preference Aided Imitation Learning from Imperfect Demonstrations (ICML 2024 · similarity 0.88)
- GO-DICE: Goal-Conditioned Option-Aware Offline Imitation Learning via Stationary Distribution Correction Estimation (AAAI 2024 · similarity 0.88)
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.
