Event

UCA-Flow Unified Condition-Action Modeling Framework Released

1 reports · 1 sources

Xinyu Zhou and co-authors propose UCA-Flow, a unified condition-action modeling framework that encodes observation conditions, timestep conditions, interval conditions, and action tokens into a single sequence, jointly learned by a unified Transformer to achieve accurate one-step action generation. The authors report that UCA-Flow achieves an average success rate of 86.4%, 7.5 percentage points higher than the strongest baseline, with an average task inference latency of 3.0 milliseconds.