Research map / Label-efficient and robust learning / Recognition and long-tail
TransHP: Image Classification with Hierarchical Prompting
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 prompting (TransHP) | light transformer baseline · ImageNet | top-1 accuracy (%): 76.21 -> 78.65 | Helped |
| supervise class token with coarse labels, no prompts (ablation) | TransHP · ImageNet | top-1 accuracy (%): +1.37% | Helped |
| inject learnable tokens without coarse-label supervision (ablation) | light transformer baseline · ImageNet | top-1 accuracy (%) | No clear effect |
| use prompt tokens instead of only coarse-label supervision (ablation) | No prompts variant · ImageNet | top-1 accuracy (%): -1.07% | Helped |
| add TransHP when training from scratch vs pre-trained | baseline · CIFAR-100 | top-1 accuracy (%): +5.32% and +1.87% | Helped |
| add TransHP to ViT-B/16 | ViT-B/16 · ImageNet | top-1 accuracy (%): 76.68 -> 79.51 | Helped |
| add TransHP to ViT-L/16 | ViT-L/16 · ImageNet | top-1 accuracy (%): 76.37 -> 78.80 | Helped |
| add TransHP to DeiT-S | DeiT-S · ImageNet | top-1 accuracy (%): 79.82 -> 80.55 | Helped |
| add TransHP to DeiT-B | DeiT-B · ImageNet | top-1 accuracy (%): 81.80 -> 82.35 | Helped |
| reduce training data to 10% | TransHP vs baseline · ImageNet 10% data | top-1 accuracy (%): 25.24 -> 37.93 | Helped |
| use TransHP instead of Guided | Guided · ImageNet | top-1 accuracy (%): 76.21 -> 76.05 | No clear effect |
| use TransHP instead of HiMulConE | HiMulConE · ImageNet | top-1 accuracy (%): 77.52 -> 78.65 | Helped |
Closest papers by meaning
- Learning Hierarchical Image Segmentation For Recognition and By Recognition (ICLR 2024 · similarity 0.85)
- AutoVP: An Automated Visual Prompting Framework and Benchmark (ICLR 2024 · similarity 0.85)
- L-WISE: Boosting Human Visual Category Learning Through Model-Based Image Selection and Enhancement (ICLR 2025 · similarity 0.84)
- Improving Visual Recognition with Hyperbolical Visual Hierarchy Mapping (CVPR 2024 · similarity 0.84)
- Prompt-CAM: Making Vision Transformers Interpretable for Fine-Grained Analysis (CVPR 2025 · similarity 0.84)
- Tree of Attributes Prompt Learning for Vision-Language Models (ICLR 2025 · similarity 0.84)
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.