用局部内在维数检测扩散模型幻觉，有效提升生成图像的结构一致性。
Local Intrinsic Dimension Unveils Hallucinations in Diffusion Models
- 通过模型诱导流形上的不稳定性识别幻觉
- 局部内在维数是幻觉的主要驱动因素，降低它可显著减少异常生成
- 适用于医学影像等需结构准确性的下游任务
扩散模型容易生成结构幻觉——即符合训练数据统计特性但违背底层结构规则的样本，例如手部多出五指。现有研究从模式插值等角度部分解释了该现象。本文提出新视角：将幻觉视为模型诱导流形上的不稳定性。我们发现基于此类不稳定的幻觉过滤器性能可媲美甚至超过近期提出的时序过滤方法。进一步分析表明，局部内在维数（LID）是主要驱动因素，并提出直接修正机制「内在抑制（IQ）」以降低LID，从而缓解幻觉。IQ在多种基准测试中持续优于标准幻觉抑制基线，为下游医学影像任务中的解剖一致性提供了极具前景的解决方案。
原文摘要 · Abstract (English)
Diffusion models are prone to generating structural hallucinations - samples that match the statistical properties of the training data yet defy underlying structural rules, resulting in anomalies like hands with more than five fingers. Recent research studied this failure mode from several viewpoints, offering partial explanations to their occurrence, such as mode interpolation. In this work, we propose a complementary perspective that treats hallucinations as instabilities on the model-induced manifold. We begin by showing that a hallucination filter based on such instabilities matches or exceeds the performance of the recently proposed temporal one. By tracing the source of these instabilities, we identify local intrinsic dimension (LID) as their primary driver and propose Intrinsic Quenching (IQ), a direct corrective mechanism that deflates it to alleviate hallucinations. IQ consistently outperforms standard hallucination reduction baselines across a wide array of benchmarks and offers a highly promising solution for enforcing anatomical consistency in downstream medical imaging tasks.
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。