ordinal-scoring

Rank-Then-Act (RTA) is a framework for learning control policies from expert video demonstrations without environment rewards, using a Vision-Language Model as a progress-based ordinal scorer with correlation-based rewards. It achieves stable cross-task transfer and outperforms prior methods on discrete and continuous control benchmarks.

This paper introduces the Gauge-Fixed Ordinal Network (GON), a temporal convolutional model that assigns consistent predictability scores across different dynamical systems by fixing the gauge freedom of ordinal scoring. The method transfers better than training from scratch on held-out systems, with zero-shot scores retaining ordinal structure at the stochastic boundary.