dorsal/arxiv
View SchemaInfoSculpt: Sculpting the Latent Space for Generalized Category Discovery
| Authors | Wenwen Liao, Hang Ruan, Jianbo Yu, Yuansong Wang, Qingchao Jiang, Xiaofeng Yang |
|---|---|
| Categories | |
| ArXiv ID | 2601.10098vv1 |
| URL | https://arxiv.org/abs/2601.10098 |
| License | http://arxiv.org/licenses/nonexclusive-distrib/1.0/ |
Abstract
Generalized Category Discovery (GCD) aims to classify instances from both known and novel categories within a large-scale unlabeled dataset, a critical yet challenging task for real-world, open-world applications. However, existing methods often rely on pseudo-labeling, or two-stage clustering, which lack a principled mechanism to explicitly disentangle essential, category-defining signals from instance-specific noise. In this paper, we address this fundamental limitation by re-framing GCD from an information-theoretic perspective, grounded in the Information Bottleneck (IB) principle. We introduce InfoSculpt, a novel framework that systematically sculpts the representation space by minimizing a dual Conditional Mutual Information (CMI) objective. InfoSculpt uniquely combines a Category-Level CMI on labeled data to learn compact and discriminative representations for known classes, and a complementary Instance-Level CMI on all data to distill invariant features by compressing augmentation-induced noise. These two objectives work synergistically at different scales to produce a disentangled and robust latent space where categorical information is preserved while noisy, instance-specific details are discarded. Extensive experiments on 8 benchmarks demonstrate that InfoSculpt validating the effectiveness of our information-theoretic approach.
{
"annotation_id": "7a5ed0b3-7e93-4641-91df-8291d1a4b8d8",
"date_created": "2026-02-17T05:53:23.271000Z",
"date_modified": "2026-02-17T05:53:23.271000Z",
"file_hash": "71507d8e25c719004f5e50d29b4e6e8132e9b021bc95c34d3ae7f7d3f0dbee83",
"private": false,
"record": {
"abstract": "Generalized Category Discovery (GCD) aims to classify instances from both known and novel categories within a large-scale unlabeled dataset, a critical yet challenging task for real-world, open-world applications. However, existing methods often rely on pseudo-labeling, or two-stage clustering, which lack a principled mechanism to explicitly disentangle essential, category-defining signals from instance-specific noise. In this paper, we address this fundamental limitation by re-framing GCD from an information-theoretic perspective, grounded in the Information Bottleneck (IB) principle. We introduce InfoSculpt, a novel framework that systematically sculpts the representation space by minimizing a dual Conditional Mutual Information (CMI) objective. InfoSculpt uniquely combines a Category-Level CMI on labeled data to learn compact and discriminative representations for known classes, and a complementary Instance-Level CMI on all data to distill invariant features by compressing augmentation-induced noise. These two objectives work synergistically at different scales to produce a disentangled and robust latent space where categorical information is preserved while noisy, instance-specific details are discarded. Extensive experiments on 8 benchmarks demonstrate that InfoSculpt validating the effectiveness of our information-theoretic approach.",
"arxiv_id": "2601.10098",
"authors": [
"Wenwen Liao",
"Hang Ruan",
"Jianbo Yu",
"Yuansong Wang",
"Qingchao Jiang",
"Xiaofeng Yang"
],
"categories": [
"cs.CV"
],
"license": "http://arxiv.org/licenses/nonexclusive-distrib/1.0/",
"title": "InfoSculpt: Sculpting the Latent Space for Generalized Category Discovery",
"url": "https://arxiv.org/abs/2601.10098",
"version": "v1"
},
"schema_id": "dorsal/arxiv",
"source": {
"execution_id": "6c3bc96e-73bd-403d-b8d4-32050f68d4a5",
"id": "arXiv Dataset",
"type": "Model",
"variant": "snapshot-2026-01-17",
"version": "0.1.0"
},
"user_id": 1000002
}