CloSeR: Unified Relational Distillation from Closed-Set Teachers for Category Discovery

• URL: http://arxiv.org/abs/2608.25692v1
• Date: Wed, 26 Aug 2026 12:10:54 GMT
• Title: CloSeR: Unified Relational Distillation from Closed-Set Teachers for Category Discovery
• Abstract summary: CloSeR injects Closed-Set knowledge into Generalized Category Discovery training.<n>Experiments with DINO and DINOv2 backbones on six benchmarks show consistent gains over GCD baselines.
• Score: 16.721682363090043
• License: http://creativecommons.org/licenses/by-nc-sa/4.0/
• Abstract: Generalized Category Discovery (GCD) is an intriguing open-world problem that has garnered increasing attention: given partially labelled data, the goal is to correctly recognize known classes while discovering coherent novel categories from unlabelled samples. Recent GCD methods typically adapt foundation models by jointly optimizing supervised classification and unsupervised discovery objectives on mixed labelled and unlabelled data. While effective, this coupled training can entangle closed-set recognition and open-set discovery, leading to objective conflict and biased predictions, and may disturb the semantic geometry of pretrained representations under limited labels and noisy pseudo-labels. We propose CloSeR, a simple plug-and-play framework that injects Closed-Set Relational knowledge into GCD training. CloSeR first builds a domain-adapted closed-set teacher by tuning lightweight block-wise adapters on labelled known-class data while keeping the foundation model backbone frozen, thereby preserving pretrained priors at low training cost. It then transfers the teacher's knowledge to downstream GCD via Unified Relational Distillation (URD), which distills complementary global sample-to-prototype relations to anchor known-class semantics and local sample-to-sample relations to preserve neighborhood structure, using separate feature pathways to reduce optimization interference. CloSeR is head-agnostic and readily integrates with both parametric and non-parametric GCD methods. Extensive experiments with DINO and DINOv2 backbones on six benchmarks (CIFAR-10/100, ImageNet-100, CUB, Stanford-Cars, and FGVC-Aircraft) show consistent gains over GCD baselines, achieving state-of-the-art performance. Project page: https://visual-ai.github.io/closer/

Related papers

• Sharpness-aware Dynamic Anchor Selection for Generalized Category Discovery [61.694524826522205]
  Given some labeled data of known classes, GCD aims to cluster unlabeled data that contain both known and unknown classes.<n>Large pre-trained models have a preference for some specific visual patterns, resulting in encoding spurious correlation for unlabeled data.<n>We propose a novel method, which contains two modules: Loss Sharpness Penalty (LSP) and Dynamic Anchor Selection (DAS)
  arXiv  Detail & Related papers  (2025-12-15T02:24:06Z)
• Generalized Category Discovery via Reciprocal Learning and Class-Wise Distribution Regularization [6.696520328216944]
  Generalized Category Discovery (GCD) aims to identify unlabeled samples by leveraging the base knowledge from labeled ones.<n>Recent parametric-based methods suffer from inferior base discrimination due to unreliable self-supervision.<n>We propose a Reciprocal Learning Framework (RLF) that introduces an auxiliary branch devoted to base classification.
  arXiv  Detail & Related papers  (2025-06-03T00:12:39Z)
• DebGCD: Debiased Learning with Distribution Guidance for Generalized Category Discovery [14.222473509173357]
  We tackle the problem of Generalized Category Discovery (GCD)<n>In GCD, an inherent label bias exists between known and unknown classes due to the lack of ground-truth labels for the latter.<n>We introduce DebGCD, a underlineDebiased learning with distribution guidance framework for underlineGCD.
  arXiv  Detail & Related papers  (2025-04-07T07:56:01Z)
• ProtoGCD: Unified and Unbiased Prototype Learning for Generalized Category Discovery [42.965641047139904]
  Generalized category discovery (GCD) is a pragmatic but underexplored problem.<n>Unlabeled data contain both old and new classes.<n>ProtoGCD achieves state-of-the-art performance on both generic and fine-grained datasets.
  arXiv  Detail & Related papers  (2025-04-02T06:13:14Z)
• Generalized Class Discovery in Instance Segmentation [7.400926717561454]
  We propose an instance-wise temperature assignment (ITA) method for contrastive learning and class-wise reliability criteria for pseudo-labels.<n>We evaluate our proposed method by conducting experiments on two settings: COCO$_half$ + LVIS and LVIS + Visual Genome.
  arXiv  Detail & Related papers  (2025-02-12T06:26:05Z)
• Happy: A Debiased Learning Framework for Continual Generalized Category Discovery [54.54153155039062]
  This paper explores the underexplored task of Continual Generalized Category Discovery (C-GCD) C-GCD aims to incrementally discover new classes from unlabeled data while maintaining the ability to recognize previously learned classes. We introduce a debiased learning framework, namely Happy, characterized by Hardness-aware prototype sampling and soft entropy regularization.
  arXiv  Detail & Related papers  (2024-10-09T04:18:51Z)
• Generalized Category Discovery with Clustering Assignment Consistency [56.92546133591019]
  Generalized category discovery (GCD) is a recently proposed open-world task. We propose a co-training-based framework that encourages clustering consistency. Our method achieves state-of-the-art performance on three generic benchmarks and three fine-grained visual recognition datasets.
  arXiv  Detail & Related papers  (2023-10-30T00:32:47Z)
• OpenGCD: Assisting Open World Recognition with Generalized Category Discovery [4.600906853436266]
  A desirable open world recognition (OWR) system requires performing three tasks. We propose OpenGCD that combines three key ideas to solve the above problems sequentially. Experiments on two standard classification benchmarks and a challenging dataset demonstrate that OpenGCD not only offers excellent compatibility but also substantially outperforms other baselines.
  arXiv  Detail & Related papers  (2023-08-14T04:10:45Z)
• Dynamic Conceptional Contrastive Learning for Generalized Category Discovery [76.82327473338734]
  Generalized category discovery (GCD) aims to automatically cluster partially labeled data. Unlabeled data contain instances that are not only from known categories of the labeled data but also from novel categories. One effective way for GCD is applying self-supervised learning to learn discriminate representation for unlabeled data. We propose a Dynamic Conceptional Contrastive Learning framework, which can effectively improve clustering accuracy.
  arXiv  Detail & Related papers  (2023-03-30T14:04:39Z)
• ProtoCon: Pseudo-label Refinement via Online Clustering and Prototypical Consistency for Efficient Semi-supervised Learning [60.57998388590556]
  ProtoCon is a novel method for confidence-based pseudo-labeling. Online nature of ProtoCon allows it to utilise the label history of the entire dataset in one training cycle. It delivers significant gains and faster convergence over state-of-the-art datasets.
  arXiv  Detail & Related papers  (2023-03-22T23:51:54Z)
• Parametric Classification for Generalized Category Discovery: A Baseline Study [70.73212959385387]
  Generalized Category Discovery (GCD) aims to discover novel categories in unlabelled datasets using knowledge learned from labelled samples. We investigate the failure of parametric classifiers, verify the effectiveness of previous design choices when high-quality supervision is available, and identify unreliable pseudo-labels as a key problem. We propose a simple yet effective parametric classification method that benefits from entropy regularisation, achieves state-of-the-art performance on multiple GCD benchmarks and shows strong robustness to unknown class numbers.
  arXiv  Detail & Related papers  (2022-11-21T18:47:11Z)
• Divide and Contrast: Source-free Domain Adaptation via Adaptive Contrastive Learning [122.62311703151215]
  Divide and Contrast (DaC) aims to connect the good ends of both worlds while bypassing their limitations. DaC divides the target data into source-like and target-specific samples, where either group of samples is treated with tailored goals. We further align the source-like domain with the target-specific samples using a memory bank-based Maximum Mean Discrepancy (MMD) loss to reduce the distribution mismatch.
  arXiv  Detail & Related papers  (2022-11-12T09:21:49Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.