Diverging Flows: Detecting Extrapolations in Conditional Generation
This addresses a critical safety issue for deploying flow models in domains like medicine and robotics, offering a robust solution to prevent silent failures.
The paper tackles the problem of extrapolation hazards in Flow Matching models, which produce plausible but incorrect outputs for off-manifold conditions, and introduces Diverging Flows to enable simultaneous conditional generation and extrapolation detection without compromising performance or speed.
The ability of Flow Matching (FM) to model complex conditional distributions has established it as the state-of-the-art for prediction tasks (e.g., robotics, weather forecasting). However, deployment in safety-critical settings is hindered by a critical extrapolation hazard: driven by smoothness biases, flow models yield plausible outputs even for off-manifold conditions, resulting in silent failures indistinguishable from valid predictions. In this work, we introduce Diverging Flows, a novel approach that enables a single model to simultaneously perform conditional generation and native extrapolation detection by structurally enforcing inefficient transport for off-manifold inputs. We evaluate our method on synthetic manifolds, cross-domain style transfer, and weather temperature forecasting, demonstrating that it achieves effective detection of extrapolations without compromising predictive fidelity or inference latency. These results establish Diverging Flows as a robust solution for trustworthy flow models, paving the way for reliable deployment in domains such as medicine, robotics, and climate science.