Contrastive Flow Map Matching

Junyu Zhang, Daochang Liu, Younghyun Kim, Jong Hwan Ko, Shichao Zhang, Chang Xu, Eunbyung Park
Proceedings of the 43rd International Conference on Machine Learning, PMLR 306:157180-157201, 2026.

Abstract

Flow map matching (FMM) enables one- and few-step sampling for diffusion-style generation, yet its performance is often hindered by the mismatch between ground-truth training transitions and model-induced flow maps. We propose Contrastive Flow Map Matching (CFMM), a principled framework that explicitly aligns FMM training with practical sampling. Our approach is motivated by a joint-KL decomposition on the reverse KL divergence, which decomposes the distributional gap into a marginal mismatch over intermediate states and a conditional mismatch in endpoint reconstruction. This analysis motivates two complementary objectives: average-velocity regression for marginal alignment and a sampling-aligned InfoNCE contrastive loss for conditional refinement. CFMM is a training-only plug-in for pre-trained FMMs, incurs no inference-time overhead, and supports training FMMs from scratch. Experiments on CIFAR-10, ImageNet, and LSUN across multiple FMM baselines demonstrate consistent improvements in fidelity and perceptual quality with only modest additional training cost.

Cite this Paper


BibTeX
@InProceedings{pmlr-v306-zhang26do, title = {Contrastive Flow Map Matching}, author = {Zhang, Junyu and Liu, Daochang and Kim, Younghyun and Ko, Jong Hwan and Zhang, Shichao and Xu, Chang and Park, Eunbyung}, booktitle = {Proceedings of the 43rd International Conference on Machine Learning}, pages = {157180--157201}, year = {2026}, editor = {Zhang, Tong and Dudik, Miroslav and Jaggi, Martin and Agarwal, Alekh and Li, Sharon and Schuurmans, Dale and Zhu, Jerry and Berkenkamp, Felix and Dong, Hanze and Bietti, Alberto}, volume = {306}, series = {Proceedings of Machine Learning Research}, month = {06--11 Jul}, publisher = {PMLR}, pdf = {https://raw.githubusercontent.com/mlresearch/v306/main/assets/zhang26do/zhang26do.pdf}, url = {https://proceedings.mlr.press/v306/zhang26do.html}, abstract = {Flow map matching (FMM) enables one- and few-step sampling for diffusion-style generation, yet its performance is often hindered by the mismatch between ground-truth training transitions and model-induced flow maps. We propose Contrastive Flow Map Matching (CFMM), a principled framework that explicitly aligns FMM training with practical sampling. Our approach is motivated by a joint-KL decomposition on the reverse KL divergence, which decomposes the distributional gap into a marginal mismatch over intermediate states and a conditional mismatch in endpoint reconstruction. This analysis motivates two complementary objectives: average-velocity regression for marginal alignment and a sampling-aligned InfoNCE contrastive loss for conditional refinement. CFMM is a training-only plug-in for pre-trained FMMs, incurs no inference-time overhead, and supports training FMMs from scratch. Experiments on CIFAR-10, ImageNet, and LSUN across multiple FMM baselines demonstrate consistent improvements in fidelity and perceptual quality with only modest additional training cost.} }
Endnote
%0 Conference Paper %T Contrastive Flow Map Matching %A Junyu Zhang %A Daochang Liu %A Younghyun Kim %A Jong Hwan Ko %A Shichao Zhang %A Chang Xu %A Eunbyung Park %B Proceedings of the 43rd International Conference on Machine Learning %C Proceedings of Machine Learning Research %D 2026 %E Tong Zhang %E Miroslav Dudik %E Martin Jaggi %E Alekh Agarwal %E Sharon Li %E Dale Schuurmans %E Jerry Zhu %E Felix Berkenkamp %E Hanze Dong %E Alberto Bietti %F pmlr-v306-zhang26do %I PMLR %P 157180--157201 %U https://proceedings.mlr.press/v306/zhang26do.html %V 306 %X Flow map matching (FMM) enables one- and few-step sampling for diffusion-style generation, yet its performance is often hindered by the mismatch between ground-truth training transitions and model-induced flow maps. We propose Contrastive Flow Map Matching (CFMM), a principled framework that explicitly aligns FMM training with practical sampling. Our approach is motivated by a joint-KL decomposition on the reverse KL divergence, which decomposes the distributional gap into a marginal mismatch over intermediate states and a conditional mismatch in endpoint reconstruction. This analysis motivates two complementary objectives: average-velocity regression for marginal alignment and a sampling-aligned InfoNCE contrastive loss for conditional refinement. CFMM is a training-only plug-in for pre-trained FMMs, incurs no inference-time overhead, and supports training FMMs from scratch. Experiments on CIFAR-10, ImageNet, and LSUN across multiple FMM baselines demonstrate consistent improvements in fidelity and perceptual quality with only modest additional training cost.
APA
Zhang, J., Liu, D., Kim, Y., Ko, J.H., Zhang, S., Xu, C. & Park, E.. (2026). Contrastive Flow Map Matching. Proceedings of the 43rd International Conference on Machine Learning, in Proceedings of Machine Learning Research 306:157180-157201 Available from https://proceedings.mlr.press/v306/zhang26do.html.

Related Material