Research map / Reinforcement learning / Offline RL
MAHALO: Unifying Offline Reinforcement Learning and Imitation Learning from Observations
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 |
|---|---|---|---|
| Use MAHALO instead of the strongest listed baseline | Offline PLfO · D4RL ILfO, IL, RLfO, RL-expert, and RL-sample tasks | normalized score: MAHALO scores range from 48.43 to 106.47; strongest baselines range from 3.41 to 105.65 | Mixed |
| Use MAHALO instead of specialized baselines | Offline PLfO · Meta-World ILfO, IL, RLfO, RL-expert, and RL-sample tasks | success rate | Mixed |
| Adversarially adapt the reward with the critic rather than fixing a pretrained reward model | Offline PLfO reward learning · Meta-World tasks, especially IL and ILfO | success rate | Helped |
| Use a fixed reward prediction model rather than MAHALO's adversarial reward | MAHALO · D4RL RL tasks | normalized score | No clear effect |
| Give zero reward to all unlabeled data | Offline RL with unlabeled data · Tabular offline RL with unlabeled data | policy performance | Hurt |
| Use a fixed discriminator reward | SMODICE · D4RL hopper and walker tasks | normalized score | Hurt |
Closest papers by meaning
- A Policy-Guided Imitation Approach for Offline Reinforcement Learning (NeurIPS 2022 · similarity 0.92)
- A Simple Solution for Offline Imitation from Observations and Examples with Possibly Incomplete Trajectories (NeurIPS 2023 · similarity 0.91)
- Offline RL with Observation Histories: Analyzing and Improving Sample Complexity (ICLR 2024 · similarity 0.91)
- SEABO: A Simple Search-Based Method for Offline Imitation Learning (ICLR 2024 · similarity 0.91)
- Imitation from Observations with Trajectory-Level Generative Embeddings (NeurIPS 2026 · similarity 0.90)
- Learning on One Mode: Addressing Multi-modality in Offline Reinforcement Learning (ICLR 2025 · 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.