Our new X account is live! Follow @wizwand_team for updates
WorkDL logo mark

Gradient Short-Circuit: Efficient Out-of-Distribution Detection via Feature Intervention

About

Out-of-Distribution (OOD) detection is critical for safely deploying deep models in open-world environments, where inputs may lie outside the training distribution. During inference on a model trained exclusively with In-Distribution (ID) data, we observe a salient gradient phenomenon: around an ID sample, the local gradient directions for "enhancing" that sample's predicted class remain relatively consistent, whereas OOD samples--unseen in training--exhibit disorganized or conflicting gradient directions in the same neighborhood. Motivated by this observation, we propose an inference-stage technique to short-circuit those feature coordinates that spurious gradients exploit to inflate OOD confidence, while leaving ID classification largely intact. To circumvent the expense of recomputing the logits after this gradient short-circuit, we further introduce a local first-order approximation that accurately captures the post-modification outputs without a second forward pass. Experiments on standard OOD benchmarks show our approach yields substantial improvements. Moreover, the method is lightweight and requires minimal changes to the standard inference pipeline, offering a practical path toward robust OOD detection in real-world applications.

Jiawei Gu, Ziyue Qiao, Zechao Li• 2025

Related benchmarks

TaskDatasetResultRank
OOD DetectionCIFAR-10 standard (test)--
17
OOD DetectionCIFAR-10 Overall (Combined Shift Regimes) Final aggregate (test)
AUROC60.78
7
OOD DetectionCIFAR-10 Adversarial FGSM, PGD, and AutoAttack averages (test)
AUROC55.93
7
OOD DetectionCIFAR-10-C Corruption average (test)
AUROC60.47
7
OOD DetectionOOD-benchmarks SVHN, LSUN, iSUN, Textures, Places365 (test)
AUROC0.7429
7
Showing 5 of 5 rows

Other info

Follow for update