Research map / Reinforcement learning / Visual and representation RL
Mixture-of-World Models: Scaling Multi-Task Reinforcement Learning with Modular Latent Dynamics
Boxuan Zhang, Weipu Zhang, Zhaohan Feng, Wei Xiao, Jian Sun, Jie Chen, Gang Wang
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 |
|---|---|---|---|
| use mixture of expert Transformers with shared Transformer and multiple VAEs instead of vanilla STORM Transformer with task embeddings (ablation) | multi-task STORM · Atari 100k (26 games) | performance | Improved |
| increase number of expert Transformers from 3 to 12 (ablation) | MoW · Atari 100k (26 games) | performance | Improved |
| increase number of VAE-predictor clusters from 3 to 12 (ablation) | MoW · Atari 100k (26 games) | performance | Improved |
| add shared Transformer (ablation) | MoW · Atari 100k (26 games) | performance | Improved |
| add expert Transformers (ablation) | MoW · Atari 100k (26 games) | performance | Improved |
| add task prediction head and loss (ablation) | MoW · Atari 100k (26 games) | performance | Improved |
| add harmonious loss weighting (ablation) | MoW · Atari 100k (26 games) | performance | Improved |
| add expert balance loss (ablation) | MoW · Atari 100k (26 games) | training stability | Improved |
| use gradient-based task clustering instead of random partition (12 clusters) (ablation) | MoW · Atari 100k (26 games) | performance | Improved |
| vary warmup length within a reasonable range (e.g., 5000 steps) (ablation) | MoW · Atari and Meta-World | final performance | No clear effect |
Closest papers by meaning
- HarmonyDream: Task Harmonization Inside World Models (ICML 2024 · similarity 0.90)
- Scaling Offline Model-Based RL via Jointly-Optimized World-Action Model Pretraining (ICLR 2025 · similarity 0.88)
- EDELINE: Enhancing Memory in Diffusion-based World Models via Linear-Time Sequence Modeling (NeurIPS 2025 · similarity 0.88)
- DyMoDreamer: World Modeling with Dynamic Modulation (NeurIPS 2025 · similarity 0.88)
- XLand-100B: A Large-Scale Multi-Task Dataset for In-Context Reinforcement Learning (ICLR 2025 · similarity 0.88)
- Seek Commonality but Preserve Differences: Dissected Dynamics Modeling for Multi-modal Visual RL (NeurIPS 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.
