Diffusion models have achieved unprecedented success in text-aligned generation, largely driven by Classifier-Free Guidance (CFG). However, standard CFG operates strictly on instantaneous gradients, omitting the intrinsic curvature of the data manifold. Recent methods like Zigzag-sampling (Z-Sampling) explicitly traverse multi-step forward-backward trajectories to probe this curvature, significantly improving semantic alignment. Yet, these explicit traversals triple the Neural Function Evaluation (NFE) cost and introduce unconstrained truncation errors from off-manifold evaluations, causing cumulative drift from the true marginal distribution. In this paper, we theoretically demonstrate that the explicit zigzag sequence is topologically reducible. We propose Implicit Z-Sampling, rigorously proving that intermediate states can be algebraically annihilated via operator dualities, physically eliminating off-manifold approximation errors. To push sampling efficiency to its theoretical lower bound, we introduce $Z^2$-Sampling (Zero-cost Zigzag Sampling). Exploiting the Probability Flow ODE's temporal coherence, $Z^2$-Sampling couples implicit algebraic collapse with a dynamically cached Temporal Semantic Surrogate. This restores the standard 2-NFE baseline without sacrificing semantic exploration. We formally prove via Backward Error Analysis that this discrete collapse inherently synthesizes a directional derivative curvature penalty. Finally, extensive evaluations demonstrate that $Z^2$-Sampling structurally shatters the performance-efficiency Pareto frontier. We validate its universal applicability across diverse architectures (U-Nets, DiTs) and modalities (image/video), establishing seamless orthogonality with advanced alignment frameworks (AYS, Diffusion-DPO).
Zigzag Diffusion Sampling: Diffusion Models Can Self-Improve via Self-Reflection
The Hong Kong University of Science and Technology (Guangzhou)(香港科技大学(广州));Mohamed bin Zayed University of Artificial Intelligence(穆罕默德·本·扎耶德人工智能大学);Baidu Inc(百度公司)
TMPO: Trajectory Matching Policy Optimization for Diverse and Efficient Diffusion Alignment
Huazhong University of Science and Technology(华中科技大学);Kuaishou Technology(快手科技);Nanyang Technological University(南洋理工大学);Wuhan University(武汉大学);Tsinghua University(清华大学)
Multi-GRPO: Multi-Group Advantage Estimation for Text-to-Image Generation with Tree-Based Trajectories and Multiple Rewards
University of Chinese Academy of Sciences(中国科学院大学);Beihang University(北航);Alibaba Group(阿里巴巴集团);Beijing Jiaotong University(北京交通大学);TikTok Inc.(TikTok公司);City University of Hong Kong(香港城市大学)
Diffusion Path Alignment for Long-Range Motion Generation and Domain Transitions
Shenzhen International Graduate School, Tsinghua University(清华大学深圳国际研究生院);Institute of Information Systems, Universität Liechtenstein(列支敦士登大学信息系统研究所);Department of Computer Science, New York University(纽约大学计算机科学系);MRC Biostatistics Unit, University of Cambridge(剑桥大学MRC生物统计单元)
Physics-Grounded Motion Forecasting via Equation Discovery for Trajectory-Guided Image-to-Video Generation
Monash University(墨尔本大学);Harbin Institute of Technology(哈尔滨工业大学);Institute for Clarity in Documentation(文档清晰研究所);Inria Paris-Rocquencourt(巴黎-罗quentcourt研究所);Rajiv Gandhi University(拉朱·甘地大学);Tsinghua University(清华大学);Palmer Research Laboratories(帕勒尔研究实验室)
Accurate Global Trajectory Alignment using Poles and Road Markings
Institute of Measurement and Control Systems, Karlsruhe Institute of Technology(卡尔斯鲁厄理工学院测量与控制研究所);FZI Research Center for Information Technology(FZI信息技术研究中心)
SimVTA: Simulation-Grounded Video-Trajectory Alignment for Driving World Action Models
Shanghai Innovation Institute(上海创新研究院);OpenDriveLab at The University of Hong Kong(香港大学OpenDriveLab);Li Auto Inc.(理想汽车);Tongji University(同济大学);Tsinghua University(清华大学)
3D Geometric Tooth Alignment Planning via Deep Reinforcement Learning
School of Software Technology, Zhejiang University(浙江大学软件学院);State Key Laboratory of CAD&CG, Zhejiang University(浙江大学CAD&CG国家重点实验室);Stomatology Hospital, Zhejiang University School of Medicine(浙江大学医学院附属口腔医院)