Research map / Vision-language and multimodal / Visual question answering
What You See is What You Read? Improving Text-Image Alignment Evaluation
Michal Yarom, Yonatan Bitton, Soravit Changpinyo, Roee Aharoni, Jonathan Herzig, Oran Lang, Eran Ofek, Idan Szpektor
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 synthetic training data (COCO t2i, COCO-Con, PickaPic-Con) to SNLI-VE fine-tuning (ablation) | PaLI · SeeTRUE avg | ROC AUC: 79.7 -> 82.9 | Improved |
| add synthetic training data (ablation) | PaLI · EditBench | ROC AUC: 65.5 -> 77.2 | Improved |
| add synthetic training data (ablation) | PaLI · COCO-Con | ROC AUC: 91.2 -> 91 | No clear effect |
| add synthetic training data (ablation) | PaLI · SNLI-VE | ROC AUC: 95.1 -> 94.2 | Hurt |
| average VQ2 and fine-tuned PaLI scores (ablation) | PaLI+Syn · SeeTRUE avg | ROC AUC: 82.9 -> 84.1 | Improved |
| yes-no predicate question (type C) instead of NLI comparison (type A) (ablation) | VQ2 · EditBench mean | ROC AUC: 70.9 -> 75.8 | Improved |
| yes-no predicate (type C) instead of VQA answer comparison (type B) (ablation) | VQ2 · EditBench mean | ROC AUC: 75.4 -> 75.8 | Improved |
| add multi-word answer candidates (ablation) | VQ2 · EditBench mean | ROC AUC: 73.9 -> 75.8 | Improved |
| VQ2 reranking instead of CLIP reranking | VQ2 · DrawBench SD 1.4 | human-labeled quality: 77.4 -> 87.1 | Improved |
| VQ2 vs best prior on Winoground group score | VQ2 · Winoground | group score: 16 -> 30.5 | Improved |
| VQ2 vs fine-tuned PaLI on COCO t2i | VQ2 · COCO t2i | ROC AUC: 83.4 vs 83.2 | No clear effect |
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.
- BLIP — BLIP: Bootstrapping Language-Image Pre-training for Unified Vision-Language Understanding and Generation (ICML 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
- Discriminative Probing and Tuning for Text-to-Image Generation (CVPR 2024 · similarity 0.91)
- Aligning Text to Image in Diffusion Models is Easier Than You Think (NeurIPS 2025 · similarity 0.89)
- Text as Partial Constraint: Core–Residual Alignment for Robust Vision–Language Learning (NeurIPS 2026 · similarity 0.89)
- AlignVLM: Bridging Vision and Language Latent Spaces for Multimodal Document Understanding (NeurIPS 2025 · similarity 0.89)
- Visual-Text Cross Alignment: Refining the Similarity Score in Vision-Language Models (ICML 2024 · similarity 0.89)
- Extract Free Dense Misalignment from CLIP (AAAI 2025 · similarity 0.88)
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.
