Backbone-Equated Diffusion OOD via Sparse Internal Snapshots
Fuente:
arXiv
Gespeichert in:
| Hauptverfasser: | , , , , , |
|---|---|
| Format: | Preprint |
| Veröffentlicht: |
2026
|
| Schlagworte: | |
| Online-Zugang: | |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
| _version_ | 1866911672130600960 |
|---|---|
| author | Rouzoumka, Yadang Alexis Pinsolle, Jean Terreaux, Eugénie Morisseau, Christèle Ovarlez, Jean-Philippe Ren, Chengfang |
| author_facet | Rouzoumka, Yadang Alexis Pinsolle, Jean Terreaux, Eugénie Morisseau, Christèle Ovarlez, Jean-Philippe Ren, Chengfang |
| contents | Fair comparison between diffusion-based OOD detectors is challenging, as conclusions can vary with backbone choice, corruption parameterization, and test-time budget. We address this issue through a Mutualized Backbone-Equated (MBE) protocol that aligns canonical corruption levels and logical test-time cost across diffusion backbones. Within this setting, we introduce Canonical Feature Snapshots (CFS), a family of detectors that probes a frozen diffusion backbone using only a tiny number of native internal activations at canonical low-noise levels. On a controlled CIFAR-scale benchmark, the strongest one-forward CFS variant is CFS(1x2), while an even smaller decoder-only variant remains highly competitive. This shows that much of the relative-OOD signal exposed by frozen diffusion backbones is concentrated in a small number of sparse internal states, rather than requiring full denoising trajectories or high-capacity downstream heads. We further provide a local diagnostic theory explaining these observations through conditional encoder-decoder complementarity, diagonal-score separation, and low-noise corruption stability. The official implementation is available at https://github.com/RouzAY/cfs-diffusion-ood/. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_11014 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | Backbone-Equated Diffusion OOD via Sparse Internal Snapshots Rouzoumka, Yadang Alexis Pinsolle, Jean Terreaux, Eugénie Morisseau, Christèle Ovarlez, Jean-Philippe Ren, Chengfang Machine Learning Artificial Intelligence Fair comparison between diffusion-based OOD detectors is challenging, as conclusions can vary with backbone choice, corruption parameterization, and test-time budget. We address this issue through a Mutualized Backbone-Equated (MBE) protocol that aligns canonical corruption levels and logical test-time cost across diffusion backbones. Within this setting, we introduce Canonical Feature Snapshots (CFS), a family of detectors that probes a frozen diffusion backbone using only a tiny number of native internal activations at canonical low-noise levels. On a controlled CIFAR-scale benchmark, the strongest one-forward CFS variant is CFS(1x2), while an even smaller decoder-only variant remains highly competitive. This shows that much of the relative-OOD signal exposed by frozen diffusion backbones is concentrated in a small number of sparse internal states, rather than requiring full denoising trajectories or high-capacity downstream heads. We further provide a local diagnostic theory explaining these observations through conditional encoder-decoder complementarity, diagonal-score separation, and low-noise corruption stability. The official implementation is available at https://github.com/RouzAY/cfs-diffusion-ood/. |
| title | Backbone-Equated Diffusion OOD via Sparse Internal Snapshots |
| topic | Machine Learning Artificial Intelligence |
| url | https://arxiv.org/abs/2605.11014 |