# Diffusion-SDPO: Safeguarded Direct Preference Optimization for Diffusion Models

arXiv: https://arxiv.org/abs/2511.03317 · HTML: https://mlsanity.com/paper/2511.03317/

## Key facts

- Cited by 7 papers in the map (data through 14 Oct 2026).
- Placed in Diffusion and flow policies for robotics, one of 14 directions in Diffusion models.
- Builds on 46 of its 52 references that are in the map.
- First posted on arXiv 5 Nov 2025.

## Placed in

- Diffusion models / [Diffusion and flow policies for robotics](https://mlsanity.com/diffusion/diffusion-and-flow-policies-for-robotics/)
- LLM alignment / RLHF / preference optimisation / [Diffusion and multimodal model alignment](https://mlsanity.com/alignment/diffusion-and-multimodal-model-alignment/)

## Builds on (most cited first, 30 of 46)

- [Learning Transferable Visual Models From Natural Language Supervision](https://mlsanity.com/paper/2103.00020/) (2021)
- [Denoising Diffusion Probabilistic Models](https://mlsanity.com/paper/2006.11239/) (2020)
- [U-Net: Convolutional Networks for Biomedical Image Segmentation](https://mlsanity.com/paper/1505.04597/) (2015)
- [High-Resolution Image Synthesis with Latent Diffusion Models](https://mlsanity.com/paper/2112.10752/) (2021)
- [Score-Based Generative Modeling through Stochastic Differential Equations](https://mlsanity.com/paper/2011.13456/) (2020)
- [Diffusion Models Beat GANs on Image Synthesis](https://mlsanity.com/paper/2105.05233/) (2021)
- [Direct Preference Optimization: Your Language Model is Secretly a Reward Model](https://mlsanity.com/paper/2305.18290/) (2023)
- [Deep Unsupervised Learning using Nonequilibrium Thermodynamics](https://mlsanity.com/paper/1503.03585/) (2015)
- [Batch Normalization: Accelerating Deep Network Training by Reducing Internal Covariate Shift](https://mlsanity.com/paper/1502.03167/) (2015)
- [Scalable Diffusion Models with Transformers](https://mlsanity.com/paper/2212.09748/) (2022)
- [Classifier-Free Diffusion Guidance](https://mlsanity.com/paper/2207.12598/) (2022)
- [Flow Matching for Generative Modeling](https://mlsanity.com/paper/2210.02747/) (2022)
- [SDXL: Improving Latent Diffusion Models for High-Resolution Image Synthesis](https://mlsanity.com/paper/2307.01952/) (2023)
- [Scaling Rectified Flow Transformers for High-Resolution Image Synthesis](https://mlsanity.com/paper/2403.03206/) (2024)
- [Deep reinforcement learning from human preferences](https://mlsanity.com/paper/1706.03741/) (2017)
- [Flow Straight and Fast: Learning to Generate and Transfer Data with Rectified Flow](https://mlsanity.com/paper/2209.03003/) (2022)
- [Elucidating the Design Space of Diffusion-Based Generative Models](https://mlsanity.com/paper/2206.00364/) (2022)
- [A Survey of Large Language Models](https://mlsanity.com/paper/2303.18223/) (2023)
- [Qwen-VL: A Versatile Vision-Language Model for Understanding, Localization, Text Reading, and Beyond](https://mlsanity.com/paper/2308.12966/) (2023)
- [Scaling Autoregressive Models for Content-Rich Text-to-Image Generation](https://mlsanity.com/paper/2206.10789/) (2022)
- [ImageReward: Learning and Evaluating Human Preferences for Text-to-Image Generation](https://mlsanity.com/paper/2304.05977/) (2023)
- [Diffusion Models in Vision: A Survey](https://mlsanity.com/paper/2209.04747/) (2022)
- [Pick-a-Pic: An Open Dataset of User Preferences for Text-to-Image Generation](https://mlsanity.com/paper/2305.01569/) (2023)
- [GenEval: An Object-Focused Framework for Evaluating Text-to-Image Alignment](https://mlsanity.com/paper/2310.11513/) (2023)
- [Human Preference Score v2: A Solid Benchmark for Evaluating Human Preferences of Text-to-Image Synthesis](https://mlsanity.com/paper/2306.09341/) (2023)
- [Diffusion Model Alignment Using Direct Preference Optimization](https://mlsanity.com/paper/2311.12908/) (2023)
- [ELLA: Equip Diffusion Models with LLM for Enhanced Semantic Alignment](https://mlsanity.com/paper/2403.05135/) (2024)
- [Step1X-Edit: A Practical Framework for General Image Editing](https://mlsanity.com/paper/2504.17761/) (2025)
- [Score-based Generative Modeling of Graphs via the System of Stochastic Differential Equations](https://mlsanity.com/paper/2202.02514/) (2022)
- [Smaug: Fixing Failure Modes of Preference Optimisation with DPO-Positive](https://mlsanity.com/paper/2402.13228/) (2024)

## Built on by (most cited first, 7 of 7)

- [Think-Then-Generate: Reasoning-Aware Text-to-Image Diffusion with LLM Encoders](https://mlsanity.com/paper/2601.10332/) (2026)
- [Ovis-Image Technical Report](https://mlsanity.com/paper/2511.22982/) (2025)
- [Linear-DPO: Linear Direct Preference Optimization for Diffusion and Flow-Matching Generative Models](https://mlsanity.com/paper/2605.21123/) (2026)
- [Agentic Planning with Reasoning for Image Styling via Offline RL](https://mlsanity.com/paper/2603.07148/) (2026)
- [Arena as Offline Reward: Efficient Fine-Grained Preference Optimization for Diffusion Models](https://mlsanity.com/paper/2605.06070/) (2026)
- [Manifold Drift in Flow Preference Optimization: A Root Cause of Reward Hacking](https://mlsanity.com/paper/2608.20011/) (2026)
- [SaTPO: Text-to-Image Safety Alignment via Transition Preference Optimization](https://mlsanity.com/paper/2610.19563/) (2026)

## Abstract (arXiv)

Text-to-image diffusion models deliver high-quality images, yet aligning them with human preferences remains challenging. We revisit diffusion-based Direct Preference Optimization (DPO) for these models and identify a critical pathology: enlarging the preference margin does not necessarily improve generation quality. In particular, the standard Diffusion-DPO objective can increase the reconstruction error of both winner and loser branches. Consequently, degradation of the less-preferred outputs can become sufficiently severe that the preferred branch is also adversely affected even as the margin grows. To address this, we introduce Diffusion-SDPO, a safeguarded update rule that preserves the winner by adaptively scaling the loser gradient according to its alignment with the winner gradient. A first-order analysis yields a closed-form scaling coefficient that guarantees the error of the preferred output is non-increasing at each optimization step. Our method is simple, model-agnostic, broadly compatible with existing DPO-style alignment frameworks and adds only marginal computational overhead. Across standard text-to-image benchmarks, Diffusion-SDPO delivers consistent gains over preference-learning baselines on automated preference, aesthetic, and prompt alignment metrics. Code is publicly available at https://github.com/AIDC-AI/Diffusion-SDPO.