Research map / Reinforcement learning / Imitation learning
LobsDICE: Offline Learning from Observation via Stationary Distribution Correction Estimation
Geon-Hyeong Kim, Jongmin Lee, Youngsoo Jang, Hongseok Yang, Kee-Eung Kim
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 |
|---|---|---|---|
| Match expert state-transition stationary distributions rather than only state stationary distributions (ablation) | LobsDICE · Offline LfO state occupancy matching example | expert behavior recovery | Improved |
| Use LobsDICE instead of OPOLO | Offline LfO · Random MDPs with stochastic transitions | TV distance | Improved |
| Use LobsDICE instead of BCO | Offline LfO · Random MDPs with stochastic transitions | TV distance | Improved |
| Use LobsDICE instead of DemoDICEfO | Offline LfO · Random MDPs with stochastic transitions | TV distance | Improved |
| Use LobsDICE instead of OPOLO | Offline LfO · Gym-MuJoCo continuous control tasks | normalized score | Improved |
| Use LobsDICE instead of IQ-Learn | Offline LfO · Gym-MuJoCo continuous control tasks | normalized score | Improved |
| Increase the proportion of non-expert data in imperfect demonstrations | BCO and DemoDICEfO · Gym-MuJoCo continuous control tasks | normalized score | Hurt |
| Increase the amount of data | BCO and DemoDICEfO · Random MDPs with stochastic transitions | TV distance | No clear effect |
| Apply state-only behavior cloning to imperfect demonstrations without expert distribution matching | BC on imperfect demonstrations · Gym-MuJoCo continuous control tasks | normalized score | Hurt |
Closest papers by meaning
- IOSTOM: Offline Imitation Learning from Observations via State Transition Occupancy Matching (NeurIPS 2025 · similarity 0.90)
- Offline Imitation from Observation via Primal Wasserstein State Occupancy Matching (ICML 2024 · similarity 0.88)
- Imitation from Observations with Trajectory-Level Generative Embeddings (NeurIPS 2026 · similarity 0.86)
- Offline RL with Observation Histories: Analyzing and Improving Sample Complexity (ICLR 2024 · similarity 0.86)
- Offline Data Enhanced On-Policy Policy Gradient with Provable Guarantees (ICLR 2024 · similarity 0.85)
- Mitigating Covariate Shift in Behavioral Cloning via Robust Stationary Distribution Correction (NeurIPS 2024 · similarity 0.85)
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.
