Research map / Video / Video segmentation and tracking
Learning to Be Uncertain: Pre-training World Models with Horizon-Calibrated Uncertainty
Shenghua Wan, Le Gan, De-Chuan Zhan
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 |
|---|---|---|---|
| add HCU loss (ablation) | HAUWM · DMC | normalized return: 0.64 -> 0.74 | Improved |
| add HCU loss (ablation) | HAUWM · MetaWorld | normalized return: 0.73 -> 0.85 | Improved |
| add HCU loss (ablation) | HAUWM · RoboDesk | normalized return: 0.55 -> 0.71 | Improved |
| lambda 10.0 instead of 1.0 (ablation) | HAUWM · DMC | normalized return: 0.74 -> 0.67 | Hurt |
| lambda 0.01 instead of 1.0 (ablation) | HAUWM · DMC | normalized return: 0.74 -> 0.70 | Hurt |
| linear instead of exponential scaling of HCU (ablation) | HAUWM · Cheetah Run | episodic return: 587 -> 626 | Improved |
| linear instead of periodic scaling of HCU (ablation) | HAUWM · Cheetah Run | episodic return: 421 -> 626 | Improved |
| ensemble size 7 instead of 1 (ablation) | HAUWM · Cheetah Run | episodic return: 544 -> 626 | Improved |
| ensemble size 7 instead of 9 (ablation) | HAUWM · Cheetah Run | episodic return: 579 -> 626 | Improved |
| ensemble size 9 instead of 7 (ablation) | HAUWM · Dial Turn | episodic return: 3225 -> 3377 | Mixed |
| Kmax = 5 instead of 1, 2 or 10 (ablation) | HAUWM · DMC, MetaWorld, RoboDesk | normalized return | Improved |
| HAUWM versus ContextWM on Push Green | HAUWM · Push Green (imitation) | normalized return: 0.69 -> 0.63 | Hurt |
Closest papers by meaning
- AdaWorld: Learning Adaptable World Models with Latent Actions (ICML 2025 · similarity 0.89)
- Reward-Free Curricula for Training Robust World Models (ICLR 2024 · similarity 0.88)
- Beyond Prediction: Steering VLM Agents with Retrospective World Modeling (NeurIPS 2026 · similarity 0.88)
- Simplifying Latent Dynamics with Softly State-Invariant World Models (NeurIPS 2024 · similarity 0.88)
- Zero-shot World Models via Search in Memory (NeurIPS 2025 · similarity 0.87)
- Towards Understanding Future: Consistency Guided Probabilistic Modeling for Action Anticipation (AAAI 2024 · 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.
