Research map / Vision-language and multimodal / CLIP and zero-shot
TaiSu: A 166M Large-scale High-Quality Dataset for Chinese Vision-Language Pre-training
Yulong Liu, Guibo Zhu, Bin Zhu, Qi Song, Guojing Ge, Haoran Chen, GuanHui Qiao, Ru Peng, Lingxiang Wu, Jinqiao 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 |
|---|---|---|---|
| filter web data by image-text retrieval (ablation) | TaiSu ViT-B · Flickr8k retrieval | Mean Recall: 53.8 -> 58.2 | Improved |
| filter web data by image-text retrieval (ablation) | TaiSu ViT-B · Flickr30k retrieval | Mean Recall: 58.7 -> 64.2 | Improved |
| filter web data by image-text retrieval (ablation) | TaiSu ViT-B · COCO-CN retrieval | Mean Recall: 63.7 -> 66.4 | Improved |
| filter web data by image-text retrieval (ablation) | TaiSu ViT-B · MUGE retrieval | Mean Recall: 48.8 -> 51.4 | Improved |
| filter web data by image-text retrieval (ablation) | TaiSu ViT-B · zero-shot classification average | R@1: 48.13 -> 50.64 | Improved |
| filter web data by image-text retrieval (ablation) | TaiSu ViT-B · EuroSAT classification | R@1: 36.0 -> 29.5 | Hurt |
| add generated captions to filtered web data (ablation) | TaiSu ViT-B · Flickr8k retrieval | Mean Recall: 58.2 -> 72.6 | Improved |
| add generated captions to filtered web data (ablation) | TaiSu ViT-B · Flickr30k retrieval | Mean Recall: 64.2 -> 77.7 | Improved |
| add generated captions to filtered web data (ablation) | TaiSu ViT-B · EuroSAT classification | R@1: 29.5 -> 52.6 | Improved |
| add generated captions to filtered web data (ablation) | TaiSu ViT-B · CIFAR10 classification | R@1: 89.2 -> 85.5 | Hurt |
| add generated captions to filtered web data (ablation) | TaiSu ViT-B · zero-shot classification average | R@1: 50.64 -> 53.10 | Improved |
| add generated captions to filtered web data (ablation) | TaiSu RN101 · zero-shot classification average | R@1: 46.63 -> 48.13 | Mixed |
Closest papers by meaning
- CVLUE: A New Benchmark Dataset for Chinese Vision-Language Understanding Evaluation (AAAI 2025 · similarity 0.90)
- Vision‑Language‑Vision Auto‑Encoder: Scalable Knowledge Distillation from Diffusion Models (NeurIPS 2025 · similarity 0.88)
- DH-Set: Improving Vision-Language Alignment with Diverse and Hybrid Set-Embeddings Learning (CVPR 2025 · similarity 0.87)
- Semantic and Expressive Variations in Image Captions Across Languages (CVPR 2025 · similarity 0.87)
- Distilling Vision-Language Models on Millions of Videos (CVPR 2024 · similarity 0.87)
- Dynamic Adapter with Semantics Disentangling for Cross-lingual Cross-modal Retrieval (AAAI 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.
