Research map / Architectures and efficiency / Pruning and efficient tuning
Efficient Adaptation of Large Vision Transformer via Adapter Re-Composing
Wei Dong, Dawei Yan, Zhijun Lin, Peng Wang
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 |
|---|---|---|---|
| Use ARC instead of the strongest reported baseline | ViT-B/16 adaptation · FGVC (5-dataset mean) | accuracy: 90.7 -> 90.1 | No clear effect |
| Use ARC instead of SSF* | ViT-B/16 adaptation · VTAB-1k (Mean Total) | accuracy: 73.1 -> 73.4 | Improved |
| Increase bottleneck dimension from 10 to 50 (ablation) | ARC · VTAB-1k | accuracy: 72.4 -> 73.4 | Improved |
| Increase bottleneck dimension from 50 to 100 or 200 (ablation) | ARC · VTAB-1k | accuracy: 73.4 -> 73.1; 73.4 -> 71.4 | Hurt |
| Use adapters before both MHA and FFN rather than only before MHA (ablation) | ARC · VTAB-1k | accuracy: 72.2 -> 73.4 | Improved |
| Move adapters after both MHA and FFN rather than before them (ablation) | ARC · VTAB-1k | accuracy: 73.4 -> 71.4 | Hurt |
| Share projections within and across layers rather than use non-sharing projections (ablation) | ARC · VTAB-1k | parameters (M): 0.98 -> 0.13 | Improved |
| Insert sequentially across layers 1–12 rather than layers 1–6 only (ablation) | ARC · VTAB-1k | accuracy: 71.5 -> 73.4 | Improved |
| Use parallel rather than sequential insertion in layers 1–12 (ablation) | ARC · VTAB-1k | accuracy: 73.4 -> 70.4 | Hurt |
Closest papers by meaning
- Time- Memory- and Parameter-Efficient Visual Adaptation (CVPR 2024 · similarity 0.89)
- Adapters Strike Back (CVPR 2024 · similarity 0.87)
- Split Adaptation for Pre-trained Vision Transformers (CVPR 2025 · similarity 0.87)
- Memory Efficient Transformer Adapter for Dense Predictions (ICLR 2025 · similarity 0.87)
- Efficient Adaptation of Pre-trained Vision Transformer via Householder Transformation (NeurIPS 2024 · similarity 0.87)
- Dynamic Tuning Towards Parameter and Inference Efficiency for ViT Adaptation (NeurIPS 2024 · similarity 0.86)
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.
