# Zigzag Diffusion Sampling: Diffusion Models Can Self-Improve via Self-Reflection

arXiv: https://arxiv.org/abs/2412.10891 · HTML: https://mlsanity.com/paper/2412.10891/

## Key facts

- Cited by 41 papers in the map (data through 14 Oct 2026).
- Placed in Image generation, text-to-image and editing, one of 14 directions in Diffusion models.
- Builds on 26 of its 57 references that are in the map.
- First posted on arXiv 14 Dec 2024.

## Placed in

- Diffusion models / [Image generation, text-to-image and editing](https://mlsanity.com/diffusion/image-generation-text-to-image-and-editing/)

## Builds on (most cited first, 26 of 26)

- [GPT-4 Technical Report](https://mlsanity.com/paper/2303.08774/) (2023)
- [Denoising Diffusion Implicit Models](https://mlsanity.com/paper/2010.02502/) (2020)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [Stable Video Diffusion: Scaling Latent Video Diffusion Models to Large Datasets](https://mlsanity.com/paper/2311.15127/) (2023)
- [AnimateDiff: Animate Your Personalized Text-to-Image Diffusion Models without Specific Tuning](https://mlsanity.com/paper/2307.04725/) (2023)
- [Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis](https://mlsanity.com/paper/2306.09341/) (2023)
- [Adversarial Diffusion Distillation](https://mlsanity.com/paper/2311.17042/) (2023)
- [SV3D: Novel Multi-view Synthesis and 3D Generation from a Single Image using Latent Video Diffusion](https://mlsanity.com/paper/2403.12008/) (2024)
- [Hunyuan-DiT: A Powerful Multi-Resolution Diffusion Transformer with Fine-Grained Chinese Understanding](https://mlsanity.com/paper/2405.08748/) (2024)
- [SDXL-Lightning: Progressive Adversarial Diffusion Distillation](https://mlsanity.com/paper/2402.13929/) (2024)
- [Improving Sample Quality of Diffusion Models Using Self-Attention Guidance](https://mlsanity.com/paper/2210.00939/) (2022)
- [CFG++: Manifold-constrained Classifier Free Guidance for Diffusion Models](https://mlsanity.com/paper/2406.08070/) (2024)
- [Align Your Steps: Optimizing Sampling Schedules in Diffusion Models](https://mlsanity.com/paper/2404.14507/) (2024)
- [FreeInit: Bridging Initialization Gap in Video Diffusion Models](https://mlsanity.com/paper/2312.07537/) (2023)
- [Golden Noise for Diffusion Models: A Learning Framework](https://mlsanity.com/paper/2411.09502/) (2024)
- [Towards Mitigating Hallucination in Large Language Models via Self-Reflection](https://mlsanity.com/paper/2310.06271/) (2023)
- [ChronoMagic-Bench: A Benchmark for Metamorphic Evaluation of Text-to-Time-lapse Video Generation](https://mlsanity.com/paper/2406.18522/) (2024)
- [Not All Noises Are Created Equally:Diffusion Noise Selection and Optimization](https://mlsanity.com/paper/2407.14041/) (2024)
- [Smoothed Energy Guidance: Guiding Diffusion Models with Reduced Energy Curvature of Attention](https://mlsanity.com/paper/2408.00760/) (2024)
- [Good Seed Makes a Good Crop: Discovering Secret Seeds in Text-to-Image Diffusion Models](https://mlsanity.com/paper/2405.14828/) (2024)
- [Text-to-Image Rectified Flow as Plug-and-Play Priors](https://mlsanity.com/paper/2406.03293/) (2024)
- [The Lottery Ticket Hypothesis in Denoising: Towards Semantic-Driven Initialization](https://mlsanity.com/paper/2312.08872/) (2023)
- [IV-Mixed Sampler: Leveraging Image Diffusion Models for Enhanced Video Synthesis](https://mlsanity.com/paper/2410.04171/) (2024)
- [Synthetic Shifts to Initial Seed Vector Exposes the Brittle Nature of Latent-Based Diffusion Models](https://mlsanity.com/paper/2312.11473/) (2023)
- [Bag of Design Choices for Inference of High-Resolution Masked Generative Transformer](https://mlsanity.com/paper/2411.10781/) (2024)

## Built on by (most cited first, 30 of 41)

- [UniEdit-Flow: Unleashing Inversion and Editing in the Era of Flow Models](https://mlsanity.com/paper/2504.13109/) (2025)
- [Taming Video Models for 3D and 4D Generation via Zero-Shot Camera Control](https://mlsanity.com/paper/2509.15130/) (2025)
- [Self-Refining Video Sampling](https://mlsanity.com/paper/2601.18577/) (2026)
- [Noise Matters: Optimizing Matching Noise for Diffusion Classifiers](https://mlsanity.com/paper/2508.11330/) (2025)
- [ConsistentRFT: Reducing Visual Hallucinations in Flow-based Reinforcement Fine-Tuning](https://mlsanity.com/paper/2602.03425/) (2026)
- [Chain of Event-Centric Causal Thought for Physically Plausible Video Generation](https://mlsanity.com/paper/2603.09094/) (2026)
- [Multimodal LLM-Guided Semantic Correction in Text-to-Image Diffusion](https://mlsanity.com/paper/2505.20053/) (2025)
- [FastLightGen: Fast and Light Video Generation with Fewer Steps and Parameters](https://mlsanity.com/paper/2603.01685/) (2026)
- [HyperAlign: Hypernetwork for Efficient Test-Time Alignment of Diffusion Models](https://mlsanity.com/paper/2601.15968/) (2026)
- [Training-Free Refinement of Flow Matching with Divergence-based Sampling](https://mlsanity.com/paper/2604.04646/) (2026)
- [Consistent Story Generation: Unlocking the Potential of Zigzag Sampling](https://mlsanity.com/paper/2506.09612/) (2025)
- [Asynchronous Denoising Diffusion Models for Aligning Text-to-Image Generation](https://mlsanity.com/paper/2510.04504/) (2025)
- [Improving Classifier-Free Guidance of Flow Matching via Manifold Projection](https://mlsanity.com/paper/2601.21892/) (2026)
- [CoRe^2: Collect, Reflect and Refine to Generate Better and Faster](https://mlsanity.com/paper/2503.09662/) (2025)
- [Refining Few-Step Text-to-Multiview Diffusion via Reinforcement Learning](https://mlsanity.com/paper/2505.20107/) (2025)
- [UtilGen: Utility-Centric Generative Data Augmentation with Dual-Level Task Adaptation](https://mlsanity.com/paper/2510.24262/) (2025)
- [Diffusion-Driven Progressive Target Manipulation for Source-Free Domain Adaptation](https://mlsanity.com/paper/2510.25279/) (2025)
- [Training-Free Representation Guidance for Diffusion Models with a Representation Alignment Projector](https://mlsanity.com/paper/2601.22468/) (2026)
- [Ctrl-Z Sampling: Scaling Diffusion Sampling with Controlled Random Zigzag Explorations](https://mlsanity.com/paper/2506.20294/) (2025)
- [VMDiff: Visual Mixing Diffusion for Limitless Cross-Object Synthesis](https://mlsanity.com/paper/2509.23605/) (2025)
- [Compositional Image Synthesis with Inference-Time Scaling](https://mlsanity.com/paper/2510.24133/) (2025)
- [Z2-Sampling: Zero-Cost Zigzag Trajectories for Semantic Alignment in Diffusion Models](https://mlsanity.com/paper/2604.23536/) (2026)
- [Ada-Diffuser: Latent-Aware Adaptive Diffusion for Decision-Making](https://mlsanity.com/paper/2605.16054/) (2026)
- [DiFA: Inference-Time Forward-Process Alignment for Diffusion Models](https://mlsanity.com/paper/2607.17972/) (2026)
- [Step Back to Move Forward: Reflection-Aware Preference Optimization for Visual Generation](https://mlsanity.com/paper/2609.04282/) (2026)
- [Taming the Tri-Space Tension: ARC-Guided Hallucination Modeling and Control for Text-to-Image Generation](https://mlsanity.com/paper/2507.04946/) (2025)
- [TeEFusion: Blending Text Embeddings to Distill Classifier-Free Guidance](https://mlsanity.com/paper/2507.18192/) (2025)
- [PC-Diffusion: Aligning Diffusion Models with Human Preferences via Preference Classifier](https://mlsanity.com/paper/2511.07806/) (2025)
- [An Iteration-Free Fixed-Point Estimator for Diffusion Inversion](https://mlsanity.com/paper/2512.08547/) (2025)
- [Learning High-Quality Initial Noise for Single-View Synthesis with Diffusion Models](https://mlsanity.com/paper/2512.16219/) (2025)

## Abstract (arXiv)

Diffusion models, the most popular generative paradigm so far, can inject conditional information into the generation path to guide the latent towards desired directions. However, existing text-to-image diffusion models often fail to maintain high image quality and high prompt-image alignment for those challenging prompts. To mitigate this issue and enhance existing pretrained diffusion models, we mainly made three contributions in this paper. First, we propose diffusion self-reflection that alternately performs denoising and inversion and demonstrate that such diffusion self-reflection can leverage the guidance gap between denoising and inversion to capture prompt-related semantic information with theoretical and empirical evidence. Second, motivated by theoretical analysis, we derive Zigzag Diffusion Sampling (Z-Sampling), a novel self-reflection-based diffusion sampling method that leverages the guidance gap between denosing and inversion to accumulate semantic information step by step along the sampling path, leading to improved sampling results. Moreover, as a plug-and-play method, Z-Sampling can be generally applied to various diffusion models (e.g., accelerated ones and Transformer-based ones) with very limited coding and computational costs. Third, our extensive experiments demonstrate that Z-Sampling can generally and significantly enhance generation quality across various benchmark datasets, diffusion models, and performance evaluation metrics. For example, DreamShaper with Z-Sampling can self-improve with the HPSv2 winning rate up to 94% over the original results. Moreover, Z-Sampling can further enhance existing diffusion models combined with other orthogonal methods, including Diffusion-DPO.