New learning method improves prediction of coherent futures in changing worlds
Operator-Mismatch Contrastive Learning for World-Relative Trajectory Coherence
Machine Learning
Summary
Predicting possible future paths in a world is tricky because the rules that govern how things change might be different each time. The authors find that existing methods often can tell if a future looks believable, but can't tell if it fits the actual rules in effect. They propose a new way to train models by showing them examples of futures that are valid but use the wrong rules, helping them learn to spot which futures truly match the current world. Their approach improves the ability to score predictions in a way that respects the true dynamics of the situation.
What this means in practice
- •For robotics teams: Score predicted motion paths using dynamics-sensitive methods to improve coherence with actual environmental rules.
- •For simulation software developers: Integrate operator-mismatch contrastive learning to refine scenario simulations that differentiate plausible but rule-incompatible futures.
Authors
One Octadion, Firmansyah Sundana, Novanto Yudistira
Abstract
World models are increasingly judged by how well they rank plausible futures, yet plausibility alone is the wrong target. The same scene can evolve under many latent rules, and a future that is smooth and physically valid under one rule can be incoherent under the rule actually in force. We formalize the resulting failure, operator-blindness, and show that it is systematically hidden by standard evaluation: a scorer can attain near-perfect generic ranking while separating current-world from other-world futures no better than chance. We introduce Operator-Mismatch Contrastive (OMC) learning, whose semantic hard negatives are valid-but-wrong-operator trajectories, and a matching protocol that measures world-relative coherence directly. On a controlled synthetic benchmark with in- and out-of-distribution operator parameters, across two trajectory-scoring backbones, and on feature-level external validation with CoPhy ballsCF, OMC raises operator-sensitive scoring far above a chance baseline that nonetheless achieves 0.99 generic top-1. We prove that the wrong-operator negative is necessary both for the protocol to detect the failure and for the objective to fix it, and position OMC as an auxiliary coherence signal (a trajectory-coherence scorer, not a generative model, planner, or control agent) that can be attached to existing world-model representations.