Research map / Probabilistic and scientific modeling / Dynamical systems
Multi Time Scale World Models
Vaisakh Shaj, Saleh GHOLAM ZADEH, Ozan Demir, Luiz Ricardo Douat, Gerhard Neumann
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 |
|---|---|---|---|
| Replace the split observable/memory latent state (H=[I,0], full covariance) with H=I and diagonal covariance (ablation) | MTS3 · Long-horizon prediction | RMSE | Hurt |
| Remove action abstractions (only observations, primitive actions and observation abstractions) (ablation) | MTS3 · Long-horizon prediction | RMSE | Hurt |
| Train for 1-step-ahead prediction without imputation (ablation) | MTS3 · Long-horizon prediction | RMSE | Hurt |
| Use small H (2, 3, 5, 10) instead of 30 (ablation) | MTS3 · Hydraulics | RMSE | Hurt |
| Use very large H (75) instead of 30 (ablation) | MTS3 · Hydraulics | RMSE | Hurt |
| Use small H (2, 3, 5, 10) instead of 75 (ablation) | MTS3 · Mobile robot | RMSE | Hurt |
| Use very large H (150) instead of 75 (ablation) | MTS3 · Mobile robot | RMSE | Hurt |
| Multi-Transformer vs MTS3 on mean prediction | Multi-Transformer · Medium Maze and Panda | RMSE | Mixed |
| MTS3 vs Multi-Transformer on uncertainty | MTS3 · Kitchen | NLL: -7.3 -> -25.74 | Improved |
| MTS3 vs Multi-Transformer on uncertainty | MTS3 · Medium Maze | NLL: -0.88 -> -0.21 | Hurt |
| MTS3 vs Multi-Transformer on uncertainty | MTS3 · Half Cheetah | NLL: 0.25 -> -2.80 | Improved |
| MTS3 vs LSTM on uncertainty | MTS3 · Panda | NLL: 7.94 -> 2.79 | Improved |
Closest papers by meaning
- Simplifying Latent Dynamics with Softly State-Invariant World Models (NeurIPS 2024 · similarity 0.86)
- Learning POMDP World Models from Observations with Language-Model Priors (NeurIPS 2026 · similarity 0.86)
- Revisiting Multi-Agent World Modeling from a Diffusion-Inspired Perspective (NeurIPS 2025 · similarity 0.86)
- WALL-E: World Alignment by NeuroSymbolic Learning improves World Model-based LLM Agents (NeurIPS 2025 · similarity 0.85)
- FACTS: A Factored State-Space Framework for World Modelling (ICLR 2025 · similarity 0.85)
- Time-Aware World Model for Adaptive Prediction and Control (ICML 2025 · 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.
