AI Paper Atlas

Research map / Label-efficient and robust learning / Recognition and long-tail

H-Sets: Hessian-Guided Discovery of Set-Level Feature Interactions in Image Classifiers

Ayushi Mehrotra, Dipkamal Bhusal, Michael Clifford, Nidhi Rastogi

CVPR 2026 · Label-efficient and robust learning › Recognition and long-tail · approach: Feature attribution methods · 1 citations (OpenAlex)

Read the paperCheck your idea against 137,217 papers

AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Measured effects reported in the paper

What was changedWhereMeasuredEffect
add SAM segmentation prior (vs none) (ablation)H-Sets (ResNet) · ImageNetROAD AOPC: 0.18 -> 0.26Improved
add SAM segmentation prior (vs none) (ablation)H-Sets (DenseNet) · CUBROAD AOPC: 0.56 -> 0.65Improved
use SAM instead of Quickshift (ablation)H-Sets (DenseNet) · CUBROAD AOPC: 0.58 -> 0.65Improved
add segmentation prior (SAM) to H-Sets (ablation)H-Sets (DenseNet) · ImageNetROAD AOPC: 0.39 -> 0.38No clear effect
seed from top-IG pixel instead of random seed (with SAM) (ablation)H-Sets (ResNet) · ImageNetROAD AOPC: 0.17 -> 0.26Improved
seed from top-IG pixel instead of random seed (with SAM) (ablation)H-Sets (DenseNet) · CUBROAD AOPC: 0.52 -> 0.65Improved
use more interaction sets (1 to 5) (ablation)H-Sets (ResNet) · CUBROAD AOPC: 0.58 -> 0.63Improved
use more interaction sets (1 to 9) (ablation)H-Sets (VGG) · ImageNetGini sparsity: 0.98 -> 0.92Hurt
increase number of features nu from 250 to 5000 (ablation)H-Sets · ImageNetGini sparsity: 0.99 -> 0.85Hurt
increase number of features nu from 250 to 5000 (ablation)H-Sets · ImageNetROAD AOPC: 0.42 -> 0.37No clear effect
vary Hessian threshold mu from 0.1 to 0.8 (ablation)H-Sets · ImageNetROAD AOPC: 0.39 -> 0.39No clear effect
H-Sets vs MoXI on faithfulnessH-Sets · ImageNet (ResNet)ROAD AOPC: 0.19 -> 0.26Improved

Closest papers by meaning

Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy