Research map / Reinforcement learning / Visual and representation RL
MrCoM: A Meta-Regularized World-Model Generalizing Across Multi-Scenarios
Xuantang Xiong, Ni Mu, Runpeng Xie, Senhao Yang, Yaqing Wang, Lexiang Wang, Yao Luan, Siyuan Li, Shuang Xu, Yiqin Yang, Bo Xu
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 context (w/o C) from the world model (ablation) | MrCoM · multi-scenario DMControl | normalized return | Hurt |
| add deterministic latent state d (ablation) | MrCoM · DMControl dynamics changes | normalized return | Improved |
| add meta-state regularization loss Ls (ablation) | MrCoM · DMControl multi-scenario | normalized return | Improved |
| add meta-value regularization Lv (ablation) | MrCoM · DMControl multi-scenario | normalized return | Improved |
| MrCoM vs baselines under varying dynamics and reward | MrCoM · DMControl Hopper/Walker/Cheetah (4 train/eval settings) | normalized return | Improved |
| MrCoM vs baselines under varying dynamics only | MrCoM · DMControl Hopper/Walker/Cheetah | normalized return | Improved |
| MrCoM vs baselines under changed observation functions | MrCoM · DMControl, Add-G/Add-D/Msk-R/Msk-F | normalized return | Mixed |
| evaluate out-of-distribution scenarios vs in-distribution | MAMBA · DMControl multi-scenario | normalized return: 47.1 -> 29.4 (Walker) | Hurt |
| MrCoM on Cheetah in-distribution (α=5, β=20) compared with DreamerV3 | MrCoM · DMControl Cheetah | normalized return: 53.7 -> 48.6 | Hurt |
Closest papers by meaning
- PWM: Policy Learning with Multi-Task World Models (ICLR 2025 · similarity 0.89)
- MAMBA: an Effective World Model Approach for Meta-Reinforcement Learning (ICLR 2024 · similarity 0.89)
- Meta-DT: Offline Meta-RL as Conditional Sequence Modeling with World Model Disentanglement (NeurIPS 2024 · similarity 0.88)
- Learning World Models for Unconstrained Goal Navigation (NeurIPS 2024 · similarity 0.88)
- Masked Diffusion Language Models are Strong and Steerable Text-Based World Models for Agentic RL (NeurIPS 2026 · similarity 0.87)
- Revisiting Multi-Agent World Modeling from a Diffusion-Inspired Perspective (NeurIPS 2025 · similarity 0.87)
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.
