Radial Tail Repair: Post-Training Correction for Diffusion and Flow Models
Abstract
Diffusion and flow-matching models may struggle to accurately represent distributional tails, especially for heavy-tailed data. Existing approaches often use heavy-tailed noise or modify the generative dynamics, requiring dedicated retraining and more difficult estimation. We take a simpler approach: rather than relearning the full distribution, we correct only the model's behavior in the extreme regime. Our method builds on the radial-angular decomposition of multivariate heavy tails, where radius and direction become asymptotically independent. This allows us to preserve the directional structure learned by the pretrained model and correct only its radial tail. We estimate the tail onset from reference and generated samples, then rescale generated extremes so that their radial survival matches the target distribution. The resulting method is a lightweight post-training correction applied to a frozen diffusion or flow-matching model, requiring neither retraining nor changes to its sampling dynamics. Experiments on challenging synthetic heavy-tailed distributions and real weather and wildfire datasets show that this simple correction substantially improves extreme-tail coverage using only a small number of fitting samples.
Related
- Risk-averse design optimization with CVaR constraints via multifidelity tail-region correction
- The high dimensional monostable reaction-diffusion equation with free boundary and radial symmetry
- Automated correction of training sets for biological image segmentation by convolutional networks
- DataCupid: Stable Matching Repair for Training Data Acquisition with Delayed Feedback
- Training data attribution in diffusion models via mirrored unlearning and noise-consistent skew
BibTeX
@misc{cherkaoui2026radial,
title = {{Radial Tail Repair: Post-Training Correction for Diffusion and Flow Models}},
author = {Hamza Cherkaoui and Hélène Halconruy and Antonio Ocello},
howpublished = {arXiv},
year = {2026},
month = oct,
doi = {10.48550/arxiv.2610.16851},
eprint = {2610.16851},
archivePrefix = {arXiv},
url = {https://arxiv.org/abs/2610.16851},
note = {CC BY 4.0},
}
RIS
TY - UNPB
TI - Radial Tail Repair: Post-Training Correction for Diffusion and Flow Models
AU - Hamza Cherkaoui
AU - Hélène Halconruy
AU - Antonio Ocello
PY - 2026
DA - 2026/10/12
PB - arXiv
DO - 10.48550/arxiv.2610.16851
UR - https://arxiv.org/abs/2610.16851
AB - Diffusion and flow-matching models may struggle to accurately represent distributional tails, especially for heavy-tailed data. Existing approaches often use heavy-tailed noise or modify the generative dynamics, requiring dedicated retraining and more difficult estimation. We take a simpler approach: rather than relearning the full distribution, we correct only the model's behavior in the extreme regime. Our method builds on the radial-angular decomposition of multivariate heavy tails, where radius and direction become asymptotically independent. This allows us to preserve the directional structure learned by the pretrained model and correct only its radial tail. We estimate the tail onset from reference and generated samples, then rescale generated extremes so that their radial survival matches the target distribution. The resulting method is a lightweight post-training correction applied to a frozen diffusion or flow-matching model, requiring neither retraining nor changes to its sampling dynamics. Experiments on challenging synthetic heavy-tailed distributions and real weather and wildfire datasets show that this simple correction substantially improves extreme-tail coverage using only a small number of fitting samples.
ER -