Research map / Reinforcement learning / Policy gradient and actor-critic
WIMLE: Uncertainty‑Aware World Models with IMLE for Sample‑Efficient Continuous Control
Mehran Aghabozorgi, Alireza Moazeni, Yanshu Zhang, Ke Li
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 uncertainty-aware per-transition weighting (vs fixed w=1.0) (ablation) | WIMLE · Humanoid-run (DMC) | return curve | Improved |
| add the learned world model (vs model-free variant) (ablation) | WIMLE · Humanoid-run (DMC) | return curve | Improved |
| increase rollout horizon from H=1 to 4 to 6 (ablation) | WIMLE · Humanoid-run (DMC) | return curve | Improved |
| increase rollout horizon from H=6 to H=8 (ablation) | WIMLE · Humanoid-run (DMC) | return curve | No clear effect |
| extend rollout horizon up to H=40 without scheduling (ablation) | WIMLE · h1-hurdle-v0 (HumanoidBench) | return curve | No clear effect |
| replace unimodal Gaussian world model with IMLE world model (ablation) | WIMLE · Humanoid-run (DMC) | return curve | Improved |
| IMLE vs Gaussian weights evolve with training (ablation) | WIMLE · Humanoid-run (DMC) | weight dynamics | Improved |
| WIMLE vs BRO | WIMLE · DMC (16 tasks) @ 1M | IQM normalized return: 0.846 -> 0.871 | Improved |
| WIMLE vs BRO | WIMLE · Dog & Humanoid (DMC) @ 1M | IQM normalized return: 0.864 -> 0.897 | Improved |
| WIMLE vs SimBAV2 | WIMLE · HumanoidBench (14 tasks) @ 1M | IQM success-normalized score: 0.799 -> 0.898 | Improved |
| WIMLE vs SimBAV2 | WIMLE · MyoSuite (10 tasks) @ 1M | IQM success rate: 0.990 -> 0.980 | Hurt |
| WIMLE vs SimBAV2 at early training | WIMLE · MyoSuite (10 tasks) @ 100k | IQM success rate: 0.724 -> 0.460 | Hurt |
Closest papers by meaning
- PWM: Policy Learning with Multi-Task World Models (ICLR 2025 · similarity 0.86)
- World Models as Reference Trajectories for Rapid Motor Adaptation (NeurIPS 2025 · similarity 0.86)
- Bigger, Regularized, Categorical: High-Capacity Value Functions are Efficient Multi-Task Learners (NeurIPS 2025 · similarity 0.85)
- Efficient Model-Based Reinforcement Learning Through Optimistic Thompson Sampling (ICLR 2025 · similarity 0.85)
- Langevin Soft Actor-Critic: Efficient Exploration through Uncertainty-Driven Critic Learning (ICLR 2025 · similarity 0.85)
- Bigger, Regularized, Optimistic: scaling for compute and sample efficient continuous control (NeurIPS 2024 · 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.
