Research map / Vision-language and multimodal / Visual question answering
Large-Scale Adversarial Training for Vision-and-Language Representation Learning
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 embedding-space adversarial pre-training and finetuning (VILLA) | UNITER-large · VQA test-std | accuracy: 74.02 -> 74.87 | Helped |
| add VILLA | UNITER-large · VCR Q->AR | accuracy: +2.9 | Helped |
| add adversarial pre-training (ablation) | UNITER-base · 6-task average | average score: 78.06 -> 78.57 | Helped |
| add adversarial finetuning (ablation) | UNITER-base · 6-task average | average score: 78.06 -> 78.88 | Helped |
| combine adversarial pre-training and finetuning (ablation) | UNITER-base · 6-task average | average score: 78.06 -> 79.21 | Helped |
| perturb both modalities instead of one (ablation) | VILLA-large · VCR val Q->AR | accuracy: 64.63 -> 65.18 | Helped |
| perturb both modalities instead of image only (ablation) | VILLA-base · VCR val Q->A | accuracy: 75.81 -> 75.54 | Hurt |
| add KL-divergence adversarial regularization (vs FreeLB) (ablation) | UNITER-base finetuning · VQA test-dev | accuracy: 72.82 -> 73.29 | Helped |
| add adversarial finetuning | LXMERT · VQA, GQA, NLVR2 | Meta-Ave.: +0.88 | Helped |
| add VILLA | UNITER-base · VQA-Rephrasings | accuracy: 64.56 -> 65.35 | Helped |
Closest papers by meaning
- VladVA: Discriminative Fine-tuning of LVLMs (CVPR 2025 · similarity 0.87)
- VQAttack: Transferable Adversarial Attacks on Visual Question Answering via Pre-trained Models (AAAI 2024 · similarity 0.87)
- From Head to Tail: Towards Balanced Representation in Large Vision-Language Models through Adaptive Data Calibration (CVPR 2025 · similarity 0.87)
- NaturalBench: Evaluating Vision-Language Models on Natural Adversarial Samples (NeurIPS 2024 · similarity 0.87)
- Prismatic VLMs: Investigating the Design Space of Visually-Conditioned Language Models (ICML 2024 · similarity 0.86)
- Enhancing Large Vision Language Models with Self-Training on Image Comprehension (NeurIPS 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.