Research map / Vision-language and multimodal / CLIP and zero-shot
ZeroNLG: Aligning and Autoencoding Domains for Zero-Shot Multimodal and Multilingual Natural Language Generation
Bang Yang, Fenglin Liu, Yuexian Zou, Xian Wu, Yaowei Wang, David Andrew Clifton
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 data corruption (masking and feature noise) in denoising language reconstruction (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 5.7 -> 29.9 | Improved |
| add cross-domain alignment pre-training (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 1.0 -> 29.9 | Improved |
| add more languages (Chinese, German, French) to training (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 27.5 -> 29.9 | Improved |
| add German and French to English and Chinese training (ablation) | ZeroNLG · Flickr30k-Zh image captioning | CIDEr: 17.5 -> 18.0 | Improved |
| replace mDistilBERT encoder with sBERT (ablation) | ZeroNLG · Flickr30k-Zh image captioning | CIDEr: 18.0 -> 20.1 | Improved |
| replace mDistilBERT encoder with XLM-R (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 29.9 -> 28.7 | Hurt |
| increase decoder layers from 3 to 24 (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 29.9 -> 34.4 | Improved |
| use news-domain WMT corpus instead of CC3M for alignment (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 29.9 -> 21.2 | Hurt |
| train on WMT then CC3M sequentially (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 29.9 -> 31.6 | Improved |
| add InfoNCE loss (lambda1=1) to MSE alignment (ablation) | ZeroNLG · MS-COCO image captioning | CIDEr: 29.9 -> 27.5 | Hurt |
| use ZeroNLG instead of CapDec with Google Translator for image-to-Chinese | zero-shot captioning · Flickr30k-Zh | CIDEr: 13.9 -> 18.0 | Improved |
| use ZeroNLG instead of CapDec for English image captioning | zero-shot captioning · MS-COCO | CIDEr: 25.0 -> 29.9 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- MAGIC — MAGIC: Multimodal relAtional Graph adversarIal inferenCe for Diverse and Unpaired Text-Based Image Captioning (AAAI 2022) · better on 1 of 1 shared result
- ZeroCap — ZeroCap: Zero-Shot Image-to-Text Generation for Visual-Semantic Arithmetic (CVPR 2022) · 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
- Mining Fine-Grained Image-Text Alignment for Zero-Shot Captioning via Text-Only Training (AAAI 2024 · similarity 0.88)
- Large Multilingual Models Pivot Zero-Shot Multimodal Learning across Languages (ICLR 2024 · similarity 0.87)
- Gen-Z: Generative Zero-Shot Text Classification with Contextualized Label Descriptions (ICLR 2024 · similarity 0.86)
- Zero-shot Referring Expression Comprehension via Structural Similarity Between Images and Captions (CVPR 2024 · similarity 0.86)
- Zero-Shot Image Captioning with Multi-type Entity Representations (AAAI 2025 · similarity 0.86)
- Learning by Correction: Efficient Tuning Task for Zero-Shot Generative Vision-Language Reasoning (CVPR 2024 · similarity 0.86)
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.
