RareCount: Interpretable Out-of-Distribution Detection in Sparse Concept Space

A label- and head-free OOD detector that flags inputs by unusually active rare concepts from a sparse autoencoder, and shows which concepts drove the call.

Abstract

Vision Foundation Models (VFMs) are increasingly deployed in critical applications where confident predictions on Out-of-Distribution (OOD) inputs can pose safety risks. Existing OOD detectors can flag unreliable inputs but do not explain what drives their decisions, thus limiting adoption in high-stakes scenarios. We introduce RareCount, an interpretable, label- and head-free OOD detector that uses Sparse autoencoders (SAEs) to extract concepts. RareCount identifies OOD samples through unusually high activation of rare concepts relative to similar samples. Operating in concept space enables score decomposition and spatial grounding of the concepts responsible for detection. On OpenOOD, RareCount is competitive with post-hoc detectors, reduces the mean far-OOD FPR95 of the only other SAE-based detector by up to 14.8 points, and generalizes across DINOv2, ViT-MAE, DINOv3, and different sparsity levels. Local conditioning improves AUROC over global rarity by 26.06 points on near-OOD and 34.75 on far-OOD, while agreement, occlusion, and patch-atypicality analyses validate concept grounding, further supported by a controlled object-level demo in mixed In-Distribution (ID)/OOD scenes.

Cite

@misc{coviello2026rarecount,
  title  = {{RareCount}: Interpretable Out-of-Distribution Detection in Sparse Concept Space},
  author = {Coviello, Antonella and Onorato, Gabriele and Bastian, Nathaniel D. and Restuccia, Francesco},
  year   = {2026},
  note   = {Preprint}
}

All publications