Saved in:
| Main Authors: | , , , , , , , , |
|---|---|
| Format: | Preprint |
| Published: |
2025
|
| Subjects: | |
| Online Access: | https://arxiv.org/abs/2512.10339 |
| Tags: |
Add Tag
No Tags, Be the first to tag this record!
|
| _version_ | 1866911737492537344 |
|---|---|
| author | Lee, Ziseok Hwang, Minyeong Lee, Wooyeol Jo, Sanghyun Ko, Jihyung Park, Young Bin Choi, Jae-Mun Yang, Eunho Kim, Kyungsu |
| author_facet | Lee, Ziseok Hwang, Minyeong Lee, Wooyeol Jo, Sanghyun Ko, Jihyung Park, Young Bin Choi, Jae-Mun Yang, Eunho Kim, Kyungsu |
| contents | Inference-time steering adapts pretrained diffusion and flow models to new tasks without retraining, often utilizing ratio-of-densities constructions that reweight time-indexed marginals with fixed exponents. We identify Marginal Path Collapse, a failure mode in which the intermediate density defined by such compositions becomes non-normalizable despite valid endpoints. This collapse can arise when composing heterogeneous experts trained with mismatched noise schedules (and/or negative exponents / partial supports). To address this, we provide (i) a sharp sufficient Path Existence Criterion that characterizes when the composed intermediate densities are mathematically well-defined, and (ii) Adaptive Path Correction with Exponents (ACE), which generalizes Feynman-Kac steering to support time-varying exponents. Our analysis reveals that ACE controls the quantile radius of the intermediate distributions, providing a theoretical mechanism for path stabilization observed in experiments. On flexible-pose scaffold decoration, a drug design task composed of de-novo, conformer, and protein-conditioned experts, ACE prevents collapse and significantly outperforms constant-exponent baselines. Furthermore, ACE improves attribute success rates in compositional image generation, establishing it as a general framework for compositional sampling. Project Page: https://ziseoklee.github.io/projects/ACE/ |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2512_10339 |
| institution | arXiv |
| publishDate | 2025 |
| record_format | arxiv |
| spellingShingle | On the Collapse of Generative Paths: A Criterion and Correction for Diffusion Steering Lee, Ziseok Hwang, Minyeong Lee, Wooyeol Jo, Sanghyun Ko, Jihyung Park, Young Bin Choi, Jae-Mun Yang, Eunho Kim, Kyungsu Artificial Intelligence Inference-time steering adapts pretrained diffusion and flow models to new tasks without retraining, often utilizing ratio-of-densities constructions that reweight time-indexed marginals with fixed exponents. We identify Marginal Path Collapse, a failure mode in which the intermediate density defined by such compositions becomes non-normalizable despite valid endpoints. This collapse can arise when composing heterogeneous experts trained with mismatched noise schedules (and/or negative exponents / partial supports). To address this, we provide (i) a sharp sufficient Path Existence Criterion that characterizes when the composed intermediate densities are mathematically well-defined, and (ii) Adaptive Path Correction with Exponents (ACE), which generalizes Feynman-Kac steering to support time-varying exponents. Our analysis reveals that ACE controls the quantile radius of the intermediate distributions, providing a theoretical mechanism for path stabilization observed in experiments. On flexible-pose scaffold decoration, a drug design task composed of de-novo, conformer, and protein-conditioned experts, ACE prevents collapse and significantly outperforms constant-exponent baselines. Furthermore, ACE improves attribute success rates in compositional image generation, establishing it as a general framework for compositional sampling. Project Page: https://ziseoklee.github.io/projects/ACE/ |
| title | On the Collapse of Generative Paths: A Criterion and Correction for Diffusion Steering |
| topic | Artificial Intelligence |
| url | https://arxiv.org/abs/2512.10339 |