Research map / Vision-language and multimodal / CLIP and zero-shot
Position-Guided Text Prompt for Vision-Language Pre-Training
Jinpeng Wang, Pan Zhou, Mike Zheng Shou, Shuicheng Yan
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 position-guided text prompts to ViLT | ViLT · MSCOCO 5K test, zero-shot image-to-text retrieval | R@1: 41.3 -> 55.1 | Improved |
| Add position-guided text prompts to BLIP | BLIP · Flickr30K 1K test, zero-shot retrieval average recall | Avg: 81.7 -> 88.4 | Improved |
| Add position-guided text prompts to BLIP | BLIP · NoCaps validation overall | CIDEr: 98.8 -> 106.0 | Improved |
| Add position-guided text prompts to ViLT | ViLT · MSRVTT 1k test, text-to-video retrieval | R@1 | Improved |
| Use an additional position-prediction pretext task rather than only the prompt (ablation) | BLIP · COCO captioning | CIDEr: 121.2 -> 123.5 | Improved |
| Use the text prompt instead of the additional pretext task (ablation) | BLIP · COCO captioning | CIDEr: 123.5 -> 127.2 | Improved |
| Mix different prompt templates rather than use the standard block prompt (ablation) | BLIP · COCO retrieval | TR@1: 73.2 -> 72.3 | Hurt |
| Add object tags, prompt, and position together rather than using object tags alone (ablation) | BLIP · COCO retrieval | TR@1: 70.2 -> 73.3 | Improved |
| Add object tags, prompt, and position together rather than omitting position (ablation) | BLIP · COCO retrieval | TR@1: 70.3 -> 73.3 | Improved |
| Add object tags, prompt, and position together rather than omitting the prompt (ablation) | BLIP · COCO retrieval | TR@1: 70.8 -> 73.3 | Improved |
| Replace Faster R-CNN tags with CLIP-similarity tags (ablation) | BLIP · COCO image-to-text retrieval | R@1: 73.3 -> 72.9 | No clear effect |
| Use the smaller ResNet101 Faster R-CNN for object tags (ablation) | BLIP · COCO image-to-text retrieval | R@1: 73.3 -> 72.7 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 2 on most shared results.
- CoCa — COCA: COllaborative CAusal Regularization for Audio-Visual Question Answering (AAAI 2023) · better on 1 of 4 shared results
- Beit-3 — BEiT: BERT Pre-Training of Image Transformers (ICLR 2022) · better on 0 of 2 shared results
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
- Your Large Vision-Language Model Only Needs A Few Attention Heads For Visual Grounding (CVPR 2025 · similarity 0.91)
- SimVG: A Simple Framework for Visual Grounding with Decoupled Multi-modal Fusion (NeurIPS 2024 · similarity 0.90)
- Investigating Compositional Challenges in Vision-Language Models for Visual Grounding (CVPR 2024 · similarity 0.90)
- Connecting the Dots: Training-Free Visual Grounding via Agentic Reasoning (AAAI 2026 · similarity 0.90)
- GroundVLP: Harnessing Zero-Shot Visual Grounding from Vision-Language Pre-training and Open-Vocabulary Object Detection (AAAI 2024 · similarity 0.90)
- LaViP: Language-Grounded Visual Prompting (AAAI 2024 · 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.
