![[2602.13061] Diverging Flows: Detecting Extrapolations in Conditional Generation](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.13061] Diverging Flows: Detecting Extrapolations in Conditional Generation
Summary
The paper introduces Diverging Flows, a method for detecting extrapolations in conditional generation models, enhancing safety in applications like robotics and climate forecasting.
Why It Matters
As AI systems are increasingly deployed in safety-critical domains, ensuring their reliability is paramount. Diverging Flows addresses the risk of silent failures in flow models, making it a significant advancement for trustworthy AI applications in fields such as medicine and robotics.
Key Takeaways
- Diverging Flows enables simultaneous conditional generation and extrapolation detection.
- The method enhances safety by addressing the critical extrapolation hazard in flow models.
- Evaluation shows effective extrapolation detection without compromising predictive performance.
Computer Science > Machine Learning arXiv:2602.13061 (cs) [Submitted on 13 Feb 2026] Title:Diverging Flows: Detecting Extrapolations in Conditional Generation Authors:Constantinos Tsakonas, Serena Ivaldi, Jean-Baptiste Mouret View a PDF of the paper titled Diverging Flows: Detecting Extrapolations in Conditional Generation, by Constantinos Tsakonas and 2 other authors View PDF Abstract: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...
