Distribution Matching Distillation without Fake Score Network
Quick Take
FSF-DMD eliminates the need for a fake-score network in distribution matching distillation for flow-map generators.
Key Points
- Proposes a pseudo-velocity surrogate for flow-map generators.
- Achieves lower FID than DMD2 in ImageNet-1K experiments.
- Includes a self-teacher variant for training from scratch.
📖 Reader Mode
~2 min readAbstract:Distribution Matching Distillation (DMD) provides an effective distribution-level correction for few-step generation, while relying on an auxiliary fake-score network to track the evolving generative distribution. Recent work combines DMD-style objectives with flow-map generators to exploit both forward-divergence training and reverse-divergence correction. The fake-score estimator remains an additional component with memory and update overhead. In this work, we study whether this explicit tracker can be avoided when the generator itself has a flow-map structure. We propose Fake-Score-network-Free DMD (FSF-DMD), a DMD formulation for flow-map generators that replaces the auxiliary fake-score estimator with a generator-induced pseudo-velocity surrogate. The key observation is that the endpoint pseudo-velocity of a flow-map generator provides a tractable proxy for fake-velocity estimation, allowing the generator itself to supply the reverse-divergence signal. Building on this observation, we derive a practical objective, extend it with flow-map-consistent backward simulation, and introduce a self-teacher variant for training from scratch. In our ImageNet-1K $256 \times 256$ experiments, FSF-DMD improves flow-map baselines, reaches lower FID than the listed DMD2 comparisons in the flow-map-initialized setting, and remains effective under flow-matching initialization and training from scratch.
| Subjects: | Computer Vision and Pattern Recognition (cs.CV) |
| Cite as: | arXiv:2605.19256 [cs.CV] |
| (or arXiv:2605.19256v1 [cs.CV] for this version) | |
| https://doi.org/10.48550/arXiv.2605.19256 arXiv-issued DOI via DataCite (pending registration) |
Submission history
From: Youngjoong Kim [view email]
[v1]
Tue, 19 May 2026 02:05:31 UTC (288 KB)
— Originally published at arxiv.org
More from arXiv cs.CV
See more →arXiv cs.CV·Jinhao Jing, Zheng Ma, Jinwei Liang, Qiannian Zhao, Shawn Chen, Jing Yang, Por Lip Yee, Prayag Tiwari, Jingjing Bai, Benyou Wang, Lewei Lu, Zhan Su
1d ago FeaturedOriginal
GeoSym127K: Scalable Symbolically-verifiable Synthesis for Multimodal Geometric Reasoning
AI Summary
GeoSym127K introduces a scalable neuro-symbolic framework for enhanced geometric reasoning in multimodal models.