Research map / Label-efficient and robust learning / Recognition and long-tail
SelEx: Self-Expertise in Fine-Grained Generalized Category Discovery
Sarah Rastegar, Mohammadreza Salehi, Yuki M Asano, Hazel Doughty, Cees Snoek
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 changed | Where | Measured | Effect |
|---|---|---|---|
| add Hierarchical Semi-Supervised K-means (HSSK) (ablation) | SelEx · CUB-200 (All) | clustering accuracy: 46.1 -> 62.6 | Improved |
| add unsupervised self-expertise loss alone (ablation) | SelEx · CUB-200 (All) | clustering accuracy: 46.1 -> 47.0 | No clear effect |
| add supervised self-expertise loss alone (ablation) | SelEx · CUB-200 (All) | clustering accuracy: 46.1 -> 55.5 | Improved |
| add L_USE to HSSK (ablation) | SelEx · CUB-200 (All) | clustering accuracy: 62.6 -> 54.7 | Hurt |
| add all three components together (HSSK + L_USE + L_SSE) (ablation) | SelEx · CUB-200 (All) | clustering accuracy: 68.7 -> 73.6 | Improved |
| increase hierarchy levels from none to 7 (ablation) | SelEx · CUB-200 (All) | clustering accuracy: 62.6 -> 73.6 | Improved |
| increase hierarchy from level 4 to level 5 (ablation) | SelEx · CUB-200 (All) | clustering accuracy: 71.0 -> 69.0 | Hurt |
| increase smoothing alpha (focus negatives within cluster) (ablation) | SelEx unsupervised self-expertise · CUB-200 | novel vs known accuracy | Mixed |
| switch backbone from DINOv1 to DINOv2 | SelEx · CUB-200 (All) | clustering accuracy: 73.6 -> 87.4 | Improved |
| use SelEx on coarse-grained datasets vs best baseline | SelEx · Coarse-grained average (All) | clustering accuracy: 87.1 vs 88.0 (SPTNet) | No clear effect |
Later papers that compared with this method
14 later accepted papers report SelEx as a baseline in their results tables; 14 of them report a better number on most of the results they share with it.
- Generalized Biomedicine Discovery (ECCV 2026 · better on 12 of 12 shared results)
- The Devil Is in Gradient Entanglement: Energy-Aware Gradient Coordinator for Robust Generalized Category Discovery (CVPR 2026 · better on 9 of 9 shared results)
- PartCo: Part-Level Correspondence Priors Enhance Category Discovery (ICML 2026 · better on 8 of 8 shared results)
- Fourier Self-Supervision for Fine-Grained Generalized Category Discovery (ECCV 2026 · better on 7 of 8 shared results)
- HVGCD:Rethinking Generalized Category Discovery through Hypothesis–Verification (ECCV 2026 · better on 8 of 8 shared results)
- Identifying Latent Concepts and Structures for Generalized Category Discovery (ICML 2026 · better on 4 of 4 shared results)
- Bures Generalized Category Discovery (ICLR 2026 · better on 4 of 4 shared results)
- CoGe-GCD: Reframing Generalized Category Discovery with Compositional Generalization (ICML 2026 · better on 3 of 3 shared results)
6 more not listed.
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SPTNet — SPTNet: An Efficient Alternative Framework for Generalized Category Discovery with Spatial Prompt Tuning (ICLR 2024) · better on 3 of 5 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Prior-Constrained Association Learning for Fine-Grained Generalized Category Discovery (AAAI 2025 · similarity 0.88)
- Unleashing the Potential of Model Bias for Generalized Category Discovery (AAAI 2025 · similarity 0.87)
- A Unified Knowledge Transfer Network for Generalized Category Discovery (AAAI 2024 · similarity 0.87)
- SEAL: Semantic-Aware Hierarchical Learning for Generalized Category Discovery (NeurIPS 2025 · similarity 0.86)
- Transfer and Alignment Network for Generalized Category Discovery (AAAI 2024 · similarity 0.86)
- Active Generalized Category Discovery (CVPR 2024 · similarity 0.85)
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.
