analiza aiqbench
This paper addresses a critical vulnerability in Flow Matching generative models by introducing robustness mechanisms against target corruption and label noise. The authors propose specific weighting strategies—Temporal Residual Score for data corruption and Comparative Conditional Score for noisy labels—to stabilize training when clean data assumptions are violated. These contributions advance the practical deployment of flow-based models in real-world scenarios where input fidelity cannot be guaranteed.
Robust Flow Matching under Target Corruption and Label Noise
De la arXiv
Text integralarXiv:2611.13324v1 Announce Type: new Abstract: Flow Matching (FM) is a strong framework for generative modeling due to its stable training and efficient sampling. However, FM assumes clean target data and labels, an assumption often violated in practice. We formulate FM training as a time-dependent regression problem over conditional vector fields and analyze how the two corruption types affect the objective: i) target corruption adds a time-dependent residual term that grows with time, ii) label noise causes updates to the wrong class-conditional vector field.
Based on this analysis, we propose to improve FM robustness by mapping corruption-specific residual scores to sample weights. For target corruption, we propose the Temporal Residual Score (TRS), which emphasizes residual magnitudes at intermediate-to-late FM times, where clean and corrupted samples are more separable. For label noise, we propose the Comparative Conditional Score (CCS), which compares the residual of the observed class-conditional vector field against alternative class-conditional vector fields for the same trajectory and velocity target.
Experiments across controlled corruption settings and real-world noisy labels show that the proposed methods improve robustness over standard FM and robust-training baselines, supporting corruption-specific residual structure as a practical basis for robust FM.
Abstract:Flow Matching (FM) is a strong framework for generative modeling due to its stable training and efficient sampling. However, FM assumes clean target data and labels, an assumption often violated in practice. We formulate FM training as a time-dependent regression problem over conditional vector fields and analyze how the two corruption types affect the objective: i) target corruption adds a time-dependent residual term that grows with time, ii) label noise causes updates to the wrong class-conditional vector field.
Based on this analysis, we propose to improve FM robustness by mapping corruption-specific residual scores to sample weights. For target corruption, we propose the Temporal Residual Score (TRS), which emphasizes residual magnitudes at intermediate-to-late FM times, where clean and corrupted samples are more separable. For label noise, we propose the Comparative Conditional Score (CCS), which compares the residual of the observed class-conditional vector field against alternative class-conditional vector fields for the same trajectory and velocity target.
Experiments across controlled corruption settings and real-world noisy labels show that the proposed methods improve robustness over standard FM and robust-training baselines, supporting corruption-specific residual structure as a practical basis for robust FM.
Accepted at NeurIPS 2026. 37 pages, 10 figures
Machine Learning (cs.LG); Computer Vision and Pattern Recognition (cs.CV)
arXiv:2611.13324 [cs.LG]
(or arXiv:2611.13324v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2611.13324 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Text integral sub licența deschisă a publicației · 295 cuvinte
Subiecte conexe
PRISM: Distribution-Gated Flow Matching for Controllable Unpaired Image Translation
arXiv · 7 aug. 2026
SCIT: Testing Causal Cache Carriers in Latent Chain-of-Thought Models
arXiv · 28 aug. 2026
One Image, No Tokens: A Controlled Study of Glyph-Based Chinese Language Modeling
arXiv · 29 sept. 2026
Controlled Acquisition and Abstention in Three-Channel Score Conflicts
arXiv · 9 oct. 2026