Research map / Vision-language and multimodal / CLIP and zero-shot
Single-Stream Multi-level Alignment for Vision-Language Pretraining
Zaid Khan, Vijay Kumar B G, Xiang Yu, Samuel Schulter, Manmohan Chandraker, Yun Fu
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 image-text contrastive loss (ITC) to ITM (ablation) | SIMLA · Flickr30k 0-shot | TR@1: 6.1 -> 73.1 | Improved |
| add masked language modeling (MLM) (ablation) | SIMLA · Flickr30k 0-shot | TR@1: 73.1 -> 84.0 | Improved |
| add masked image modeling (MIM) (ablation) | SIMLA · Flickr30k 0-shot | TR@1: 84.0 -> 85.1 | Improved |
| add pseudo-label supervision (PLS) (ablation) | SIMLA · Flickr30k 0-shot | TR@1: 85.1 -> 86.2 | Improved |
| add pseudo-label supervision (PLS) (ablation) | SIMLA · Flickr30k 0-shot | IR@1: 70.1 -> 69.5 | No clear effect |
| initialize from pretrained BERT/DeiT instead of scratch (ablation) | SIMLA (591k pairs) · Flickr 0-shot | TR@1: 18.8 -> 61.0 | Improved |
| initialize from pretrained BERT/DeiT instead of scratch (ablation) | SIMLA (591k pairs) · RefCOCO+ | TestA: 18.4 -> 36.8 | Improved |
| use SIMLA instead of ALBEF | SIMLA · Flickr30K zero-shot | TR R@1: 90.5 -> 91.9 | Improved |
| use SIMLA instead of ALBEF | SIMLA · MSCOCO 5K fine-tuned | TR R@1: 73.1 -> 75.8 | Improved |
| use SIMLA instead of ALBEF | SIMLA · VQA test-dev | accuracy: 74.5 -> 74.5 | No clear effect |
| use SIMLA instead of ALBEF | SIMLA · RefCOCO+ TestA | accuracy: 65.9 -> 67.4 | Improved |
| use SIMLA instead of KaleidoBERT | SIMLA · FashionGen | TR@1: 33.8 -> 48.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- KaleidoBERT — Kaleido-BERT: Vision-Language Pre-Training on Fashion Domain (CVPR 2021) · 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
- Text as Partial Constraint: Core–Residual Alignment for Robust Vision–Language Learning (NeurIPS 2026 · similarity 0.92)
- Analyzing Fine-Grained Alignment and Enhancing Vision Understanding in Multimodal Language Models (NeurIPS 2025 · similarity 0.92)
- DINOv2 Meets Text: A Unified Framework for Image- and Pixel-Level Vision-Language Alignment (CVPR 2025 · similarity 0.91)
- Fine-Grained Verifiers: Preference Modeling as Next-token Prediction in Vision-Language Alignment (ICLR 2025 · similarity 0.91)
- Improving fine-grained understanding in image-text pre-training (ICML 2024 · similarity 0.91)
- SmartCLIP: Modular Vision-language Alignment with Identification Guarantees (CVPR 2025 · similarity 0.90)
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.
