Research map / Vision-language and multimodal / Human and embodied modeling
VL-BERT: Pre-training of Generic Visual-Linguistic Representations
Weijie Su, Xizhou Zhu, Yue Cao, Bin Li, Lewei Lu, Furu Wei, Jifeng Dai
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 |
|---|---|---|---|
| pre-train VL-BERT on visual-linguistic and text-only data (ablation) | VL-BERT_BASE · VCR Q→AR | accuracy: 54.2 -> 55.2 | Improved |
| pre-train VL-BERT on visual-linguistic and text-only data (ablation) | VL-BERT_BASE · VQA v2.0 test-dev | accuracy: 69.58 -> 71.16 | Improved |
| pre-train VL-BERT on visual-linguistic and text-only data (ablation) | VL-BERT_BASE · RefCOCO+ | accuracy: 74.41 -> 79.88 | Improved |
| use masked language modeling with visual clues (ablation) | VL-BERT pre-training · VCR Q→A val (0.5× fine-tuning epochs) | accuracy: 72.9 -> 72.9 | No clear effect |
| add masked RoI classification with linguistic clues (ablation) | VL-BERT pre-training · VQA test-dev (0.5× fine-tuning epochs) | accuracy: 71.0 -> 71.1 | No clear effect |
| add masked RoI classification with linguistic clues (ablation) | VL-BERT pre-training · RefCOCO+ detected regions val (0.5× fine-tuning epochs) | accuracy: 69.1 -> 70.7 | Improved |
| add sentence-image relationship prediction (ablation) | VL-BERT pre-training · VQA test-dev (0.5× fine-tuning epochs) | accuracy: 71.1 -> 70.3 | Hurt |
| add sentence-image relationship prediction (ablation) | VL-BERT pre-training · VCR Q→A val (0.5× fine-tuning epochs) | accuracy: 73.0 -> 72.2 | Hurt |
| add text-only corpus during pre-training (ablation) | VL-BERT pre-training · VCR QA→R val (0.5× fine-tuning epochs) | accuracy: 73.1 -> 73.8 | Improved |
| add text-only corpus during pre-training (ablation) | VL-BERT pre-training · VCR Q→A val (0.5× fine-tuning epochs) | accuracy: 73.0 -> 73.4 | Improved |
| fine-tune Fast R-CNN visual feature parameters end-to-end (ablation) | VL-BERT_BASE · VQA test-dev (0.5× fine-tuning epochs) | accuracy: 71.1 -> 71.2 | Improved |
Later papers that compared with this method
1 later accepted paper reports VL-BERT (Large) as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- UNITER: UNiversal Image-TExt Representation Learning (ECCV 2020 · 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
- VILA: On Pre-training for Visual Language Models (CVPR 2024 · similarity 0.88)
- Revisiting the Role of Language Priors in Vision-Language Models (ICML 2024 · similarity 0.87)
- The Neglected Tails in Vision-Language Models (CVPR 2024 · similarity 0.87)
- Mono-InternVL: Pushing the Boundaries of Monolithic Multimodal Large Language Models with Endogenous Visual Pre-training (CVPR 2025 · similarity 0.87)
- Probing Visual Language Priors in VLMs (ICML 2025 · similarity 0.87)
- VL-ICL Bench: The Devil in the Details of Multimodal In-Context Learning (ICLR 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.
