Plausible Videos Can Lead to Poor Decisions: Diagnosing and CURE-ing World Models for Planning
Organizations: Cornell University · NYU Grossman School of Medicine
Abstract
Action-conditioned world models are increasingly used as world models for planning, yet a world model that makes plausible, realistic-looking videos that are consistent with the training data can produce poor decisions. We address two key questions: (1) when can we trust a world model for planning, and (2) what should we do when we cannot? We introduce CURE (Certify, Upweight, Rollout, Explore), a certificate-guided technique that assesses whether a world model can be trusted for planning and determines what to do next: whether to upweight the training data to debias the model; roll out more samples at test time; or explore by collecting new data where support (how well the training data cover the planner's decisions) is missing. We deploy CURE across multiple domains, spanning simulated physics based on PhyWorld, MuJoCo arm control, and Language-Table pushing in simulation and on real videos. The decisions our certificate trusts have lower regret than those it does not (e.g., 0.06 vs. 0.38 on Language-Table sim and 0.07 vs. 0.38 on MuJoCo contact), whereas video perceptual quality does not identify good decisions. By focusing on decision quality rather than perceptual quality, CURE lowers regret over the best pessimistic decision rule by 10-25% across support levels, and over world-action models, search and policy learning by 41-64%, given the same data.
Figures & tables
| low | partial | high | |
| One fix alone (no certificate) | |||
| Standard (no fix) | 0.693 0.024 | 0.257 0.016 | 0.126 0.007 |
| Upweight only | 0.626 0.040 | 0.253 0.063 | 0.173 0.050 |
| Rollout only | 0.646 0.024 | 0.240 0.016 | 0.115 0.006 |
| Explore only | 0.436 0.019 | 0.257 0.016 | 0.126 0.007 |
| Ablation: Cure without one step | |||
World Model-Based Planning
World model-based planning uses models of how environments change to compare imagined actions and choose goal-directed behavior.
See all 294 papers