AI Paper Atlas

Research map / Continual learning / Class-incremental

Hierarchical Divide-and-Conquer Grouping for Classification Adaptation of Pre-Trained Models

Ziqian Lu, Yunlong Yu, Qinyue Tong, Jun Liu

ICCV 2025 · Continual learning › Class-incremental · approach: Exemplar-efficient class-incremental learning · 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 depth-2 recursive intra-domain grouping to depth-1 HDG-CLIP (ablation)HDG-CLIP · CUBHarmonic mean H (%): 74.5 -> 78.2Improved
Add depth-3 grouping beyond depth 2 (ablation)HDG-CLIP · AwA2Harmonic mean H (%): 94.3 -> 91.9Hurt
Fine-tune subspace-specific linear layers instead of frozen (ablation)HDG-CLIP (Depth=1) · SUNHarmonic mean H (%): 54.0 -> 75.3Improved
Apply inter-domain grouping (HDG) to frozen CLIP without tuning (ablation)CLIP · CUBHarmonic mean H (%): 55.0 -> 63.5Improved
Apply HDG on top of PSVMA featuresPSVMA · SUNHarmonic mean H (%): 52.3 -> 59.7Improved
Use similarity-based training-free grouping instead of supervised MLP or Xgboost domain classifier (ablation)HDG · AwA2AccC_N (%): 21.5 -> 98.8 (vs MLP)Improved
Add image-level distance to the domain score (ablation)HDG · SUNAccC_N (%): 92.2 -> 94.1Improved
Add LLM-guided description generation for synthetic novel-class images (ablation)HDG-CLIP · CUBHarmonic mean H (%): 6.3% reduction in H without LLMsImproved
Add concept selection filtering of generated samples (ablation)HDG-CLIP · CUBHarmonic mean H (%): +1.5% HImproved
Increase number of generated images per class beyond 20 (ablation)HDG-CLIP · CUB, AwA2, SUNHarmonic mean H (%)No clear effect
Swap text-to-image generator (GauGAN, GLIDE, DALL-E, SDXL) (ablation)HDG-CLIP · AwA2Harmonic mean H (%)No clear effect
Fine-tune CLIP on AwA2 without grouping, test on CUBCLIP · AwA2 to CUBAccH (%): 63.8 -> 17.1Hurt

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