用优化视角改进扩散采样，有效识别并减少生成模型幻觉。
RODS: Robust Optimization Inspired Diffusion Sampling for Detecting and Reducing Hallucination in Generative Models
- 将扩散采样重构成优化问题，利用损失曲面几何特征检测高风险步骤。
- 在多个数据集上检测超70%幻觉样本，修正率超25%，不引入新伪影。
- 无需重训练，推理成本几乎不变，适合部署在现有生成系统中。
扩散模型在生成建模中达到顶尖性能，但其采样过程仍易产生幻觉，常源于得分近似不准确。本文从优化视角重新审视扩散采样，提出RODS（鲁棒优化启发的扩散采样器），通过损失曲面的几何线索检测并修正高风险采样步骤。RODS强化了更平滑的采样轨迹，并自适应调整扰动，在无需重训练且额外推理开销极低的前提下，显著降低幻觉。在AFHQv2、FFHQ和11k-hands数据集上的实验表明，RODS保持了相近的图像质量与生成多样性，同时提升了采样保真度与鲁棒性：检测超过70%的幻觉样本，修正率超过25%，且未引入新伪影。代码已开源。
原文摘要 · Abstract (English)
Diffusion models have achieved state-of-the-art performance in generative modeling, yet their sampling procedures remain vulnerable to hallucinations-often stemming from inaccuracies in score approximation. In this work, we reinterpret diffusion sampling through the lens of optimization and introduce RODS (Robust Optimization-inspired Diffusion Sampler), a novel method that detects and corrects high-risk sampling steps using geometric cues from the loss landscape. RODS enforces smoother sampling trajectories and adaptively adjusts perturbations, reducing hallucinations without retraining and at minimal additional inference cost. Experiments on AFHQv2, FFHQ, and 11k-hands demonstrate that RODS maintains comparable image quality and preserves generation diversity. More importantly, it improves both sampling fidelity and robustness, detecting over 70% of hallucinated samples and correcting more than 25%, all while avoiding the introduction of new artifacts. We release our code at https://github.com/Yiqi-Verna-Tian/RODS.
Thank you to arXiv for use of its open access interoperability. PaperDance 不是 arXiv 官方产品；中文卡片由大模型生成，请以原文为准。