Research map / Label-efficient and robust learning / Recognition and long-tail
Foundation-Adaptive Integrated Refinement for Generalized Category Discovery
Yuwei Bian, Shidong Wang, Yazhou Yao, Haofeng Zhang
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 prototype generation (hierarchical-clustering centers as classifier init) (ablation) | SimGCD-style baseline · CIFAR-100 | clustering accuracy (All): 80.1 -> 81.4 | Improved |
| add prototype generation (ablation) | SimGCD-style baseline · CUB-200 | clustering accuracy (All): 60.3 -> 65.1 | Improved |
| use adaptive fine-tuning without prototype generation (random prototypes) (ablation) | SimGCD-style baseline · CIFAR-100 | clustering accuracy (All): 80.1 -> 67.3 | Hurt |
| use adaptive fine-tuning without prototype generation (random prototypes) (ablation) | SimGCD-style baseline · CUB-200 | clustering accuracy (All): 60.3 -> 49.4 | Hurt |
| add positive cosine perturbation to estimated class count (ablation) | FAIR with prototype generation · CIFAR-100 | clustering accuracy (All): 81.4 -> 81.6 | Improved |
| add positive cosine perturbation to estimated class count (ablation) | FAIR with prototype generation · CUB-200 | clustering accuracy (All): 65.1 -> 66.0 | Improved |
| add prototype generation on top of adaptive fine-tuning (ablation) | FAIR · CUB-200 | clustering accuracy (All): 49.4 -> 71.7 | Improved |
| add weighted self-distillation (ablation) | FAIR · CIFAR-100 | clustering accuracy (All): 83.2 -> 84.2 | Improved |
| add weighted self-distillation (ablation) | FAIR · CUB-200 | clustering accuracy (All): 71.7 -> 73.7 | Improved |
| increase cosine perturbation fluctuation range A too far (ablation) | FAIR · CIFAR-100 and CUB-200 | clustering accuracy | Hurt |
| vary cosine cycle speed tau_A from 1 to 7 (ablation) | FAIR · CIFAR-100 / CUB-200 | clustering accuracy: almost 1% on CUB-200 | Mixed |
| estimate class count with FAIR instead of assuming it is known | FAIR · CUB-200 | clustering accuracy (All): 73.7 -> 72.9 | Mixed |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- selEx — SelEx: Self-Expertise in Fine-Grained Generalized Category Discovery (ECCV 2024) · better on 3 of 3 shared results
- ProtoGCD — ProtoGCD: Unified and Unbiased Prototype Learning for Generalized Category Discovery (TPAMI 2025) · better on 1 of 1 shared result
- PromptCAL — PromptCAL: Contrastive Affinity Learning via Auxiliary Prompts for Generalized Novel Category Discovery (CVPR 2023) · better on 0 of 1 shared result
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
- Generalized Category Discovery under Domain Shift: A Frequency Domain Perspective (NeurIPS 2025 · similarity 0.89)
- Transfer and Alignment Network for Generalized Category Discovery (AAAI 2024 · similarity 0.89)
- When Domain Generalization meets Generalized Category Discovery: An Adaptive Task-Arithmetic Driven Approach (CVPR 2025 · similarity 0.88)
- Active Generalized Category Discovery (CVPR 2024 · similarity 0.88)
- Prior-Constrained Association Learning for Fine-Grained Generalized Category Discovery (AAAI 2025 · similarity 0.88)
- Continual Generalized Category Discovery: Learning and Forgetting from a Bayesian Perspective (ICML 2025 · similarity 0.88)
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.
