paper-with-me

홈 › Papers

Optimizing Few-Step Generation with Adaptive Matching Distillation

2026-02-07 · Lichen Bai, Zikai Zhou, Shitong Shao, Wenliang Zhong, Shuo Yang, Shuo Chen, Bojun Chen, Zeke Xie arxiv

Distribution Matching Distillation (DMD) is a powerful acceleration paradigm, yet its stability is often compromised in Forbidden Zone, regions where the real teacher provides unreliable guidance while the fake teacher exerts insufficient repulsive force. In this work, we propose a unified optimization framework that reinterprets prior art as implicit strategies to avoid these corrupted regions. Based on this insight, we introduce Adaptive Matching Distillation (AMD), a self-correcting mechanism that utilizes reward proxies to explicitly detect and escape Forbidden Zones. AMD dynamically prioritizes corrective gradients via structural signal decomposition and introduces Repulsive Landscape Sharpening to enforce steep energy barriers against failure mode collapse. Extensive experiments across image and video generation tasks (e.g., SDXL, Wan2.1) and rigorous benchmarks (e.g., VBench, GenEval) demonstrate that AMD significantly enhances sample fidelity and training robustness. For instance, AMD improves the HPSv2 score on SDXL from 30.64 to 31.25, outperforming state-of-the-art baselines. These findings validate that explicitly rectifying optimization trajectories within Forbidden Zones is essential for pushing the performance ceiling of few-step generative models.

📄 PDF Abstract BibTeX arXiv:2602.07345

Code (0)

등록된 구현이 없습니다.

Tasks

Video Generation

Similar Papers 제목 키워드 기반

Diversity-Preserved Distribution Matching Distillation for Fast Visual Synthesis

2026-02-03 · Tianhe Wu, Ruibin Li, Lei Zhang, Kede Ma arxiv

Distribution matching distillation (DMD) facilitates few-step image generation by aligning a distilled student with a reference multi-step teacher. In practice, however, optimizing DMD can reduce sample diversity in few-…

Image Generation

Guiding Distribution Matching Distillation with Gradient-Based Reinforcement Learning

2026-04-21 · Linwei Dong, Ruoyu Guo, Ge Bai, Zehuan Yuan 외 arxiv

Diffusion distillation, exemplified by Distribution Matching Distillation (DMD), has shown great promise in few-step generation but often sacrifices quality for sampling speed. While integrating Reinforcement Learning (R…

Reinforcement Learning

Distribution Matching Distillation Meets Reinforcement Learning

2025-11-17 · Dengyang Jiang, Dongyang Liu, Zanyi Wang, Qilong Wu 외 arxiv

Distribution Matching Distillation (DMD) facilitates efficient inference by distilling multi-step diffusion models into few-step variants. Concurrently, Reinforcement Learning (RL) has emerged as a vital tool for alignin…

Reinforcement Learning

Curvature-Adaptive Consistency Flow Matching: Autonomous Trajectory Optimization via Reinforcement Learning

2026-06-21 · Songtao Tian, Guhan Chen, Bohan Li, Jingyi Ma 외 arxiv

Consistency distillation has significantly accelerated the inference of diffusion models. In this work, we reveal an intriguing asymmetry: while Logit-Normal sampling priors are highly efficacious for standard iterative …

Reinforcement Learning

Few-Step Diffusion via Score identity Distillation

2025-05-19 · Mingyuan Zhou, Yi Gu, Zhendong Wang

Diffusion distillation has emerged as a promising strategy for accelerating text-to-image (T2I) diffusion models by distilling a pretrained score network into a one- or few-step generator. While existing methods have mad…

Diversity