REVIEW 8 cited by
Extremely Simple Activation Shaping for Out-of-Distribution Detection
Not yet reviewed by Pith; the record is open.
This paper has not been read by Pith yet. Machine review is queued; the pith claim, tier, and objections will appear here once it completes.
SPECIMEN: schema-true, not a live event
T0 review · schema-true
One-sentence machine reading of the paper's core claim.
pith:XXXXXXXX · record.json · timestamp
read the original abstract
The separation between training and deployment of machine learning models implies that not all scenarios encountered in deployment can be anticipated during training, and therefore relying solely on advancements in training has its limits. Out-of-distribution (OOD) detection is an important area that stress-tests a model's ability to handle unseen situations: Do models know when they don't know? Existing OOD detection methods either incur extra training steps, additional data or make nontrivial modifications to the trained network. In contrast, in this work, we propose an extremely simple, post-hoc, on-the-fly activation shaping method, ASH, where a large portion (e.g. 90%) of a sample's activation at a late layer is removed, and the rest (e.g. 10%) simplified or lightly adjusted. The shaping is applied at inference time, and does not require any statistics calculated from training data. Experiments show that such a simple treatment enhances in-distribution and out-of-distribution distinction so as to allow state-of-the-art OOD detection on ImageNet, and does not noticeably deteriorate the in-distribution accuracy. Video, animation and code can be found at: https://andrijazz.github.io/ash
Forward citations
Cited by 8 Pith papers
-
Representation Trajectories Matters: Complementary Evidence for OOD Detection and Image Classification
Recording how an image's representation evolves block-by-block, relative to learned class routes, improves OOD detection in 131/152 comparisons and clean classification in 71/72 model–dataset cases.
-
Synthesizing Near-Boundary OOD Samples for Out-of-Distribution Detection
SynOOD generates synthetic near-boundary OOD images with MLLM-guided inpainting and energy-score gradients, then fine-tunes CLIP image and text features, reporting state-of-the-art OOD detection on ImageNet benchmarks.
-
Gradient Short-Circuit: Efficient Out-of-Distribution Detection via Feature Intervention
Gradient Short-Circuit masks the top-gradient feature coordinates, approximates the resulting logits with a first-order Taylor step, and reports large FPR95 improvements on standard OOD benchmarks.
-
A Variational Information Theoretic Approach to Out-of-Distribution Detection
A variational loss combining KL divergence and the Information Bottleneck predicts a piecewise-linear shaping function for OOD detection that beats existing element-wise shaping methods on ImageNet and CIFAR benchmarks.
-
FindMeIfYouCan: Bringing Open Set metrics to $\textit{near} $, $ \textit{far} $ and $\textit{farther}$ Out-of-Distribution Object Detection
A new OOD object detection benchmark with near, far, and farther splits and open-set metrics shows close unknown objects are found more often but are also more frequently mistaken for known objects.
-
Multi-Method Ensemble for Out-of-Distribution Detection
MME, a product of SCALE, VRA, fDBD, PCA, ViM, NME+ and CO+ scores, shows state-of-the-art OOD detection on common benchmarks, with a theoretical guarantee that is weaker than claimed.
-
Knowledge Regularized Negative Feature Tuning of Vision-Language Models for Out-of-Distribution Detection
KR-NFT tunes CLIP text features with image-conditioned scaling and shifting plus a knowledge regularization loss, improving OOD detection on base and unseen classes without forgetting pre-trained knowledge.
-
Feature Bank Enhancement for Distance-based Out-of-Distribution Detection
Clipping per-dimension outlier features in the training feature bank improves distance-based out-of-distribution detection on ImageNet-1k and CIFAR-10.
Discussion (0). Continue with ORCID to comment.