Research map / Label-efficient and robust learning / Recognition and long-tail
VL-LTR: Learning Class-Wise Visual-Linguistic Representation for Long-Tailed Visual Recognition
Changyao Tian, Wenhai Wang, Xizhou Zhu, Jifeng Dai, Yu Qiao
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 CVLP class-wise pre-training (with distillation) (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 62.8 -> 70.1 | Improved |
| initialize with CLIP pre-trained weights (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 46.8 -> 70.1 | Improved |
| add distillation loss from CLIP in pre-training (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 66.2 -> 70.1 | Improved |
| use LGR head instead of FC head (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 62.1 -> 70.1 | Improved |
| use LGR head instead of KNN head (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 63.9 -> 70.1 | Improved |
| use anchor sentence selection instead of Cut Off (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 69.7 -> 70.1 | Improved |
| add logits distillation (lambda 0.5) instead of feature distillation (lambda 0.5) (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 68.0 -> 70.1 | Improved |
| raise logits distillation weight lambda from 0.1 to 0.5 (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 68.3 -> 70.1 | Improved |
| add Wikipedia text to prompt-only descriptions (ablation) | VL-LTR (ResNet-50) · ImageNet-LT | overall top-1 accuracy (%): 69.4 -> 70.1 | Improved |
| add Wikipedia text to prompt-only descriptions (ablation) | VL-LTR (ResNet-50) · Places-LT | overall top-1 accuracy (%): 47.3 -> 48.0 | Improved |
| add class-level text and CVLP/LGR to vision-only baseline (few-shot) | Baseline (ResNet-50) · ImageNet-LT | few-shot accuracy (%): 34.5 -> 50.8 | Improved |
| add VL-LTR to vision-only baseline | Baseline (ResNet-50) · Places-LT | overall top-1 accuracy (%): 39.7 -> 48.0 | Improved |
Later papers that compared with this method
3 later accepted papers report VL-LTR as a baseline in their results tables; 2 of them report a better number on most of the results they share with it.
- CUE: Concept-Aware Multi-Label Expansion to Mitigate Concept Confusion in Long-Tailed Learning (CVPR 2026 · better on 2 of 2 shared results)
- LPT: Long-tailed Prompt Tuning for Image Classification (ICLR 2023 · better on 0 of 1 shared result)
- Long-Tailed Recognition by Mutual Information Maximization between Latent Features and Ground-Truth Labels (ICML 2023 · better on 1 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
- Vision-Language Models for Vision Tasks: A Survey (TPAMI 2024 · similarity 0.90)
- Why are Visually-Grounded Language Models Bad at Image Classification? (NeurIPS 2024 · similarity 0.88)
- Analyzing and Boosting the Power of Fine-Grained Visual Recognition for Multi-modal Large Language Models (ICLR 2025 · similarity 0.88)
- The Neglected Tails in Vision-Language Models (CVPR 2024 · similarity 0.88)
- LTGC: Long-tail Recognition via Leveraging LLMs-driven Generated Content (CVPR 2024 · similarity 0.87)
- VladVA: Discriminative Fine-tuning of LVLMs (CVPR 2025 · similarity 0.87)
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.
