AI Paper Atlas

Research map / Vision-language and multimodal / Visual question answering

A Structured, Tagged, and Localized Visual Question Answering Dataset with Full Sentence Answers and Scene Graphs for Chest X-ray Images

Philip Müller, Friederike Jungmann, Georgios Kaissis, Daniel Rueckert

ICLR 2026 · Vision-language and multimodal › Visual question answering · approach: Attention and prior-robust VQA

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 changedWhereMeasuredEffect
fine-tune on 1M fine-tuning-grade QA pairs instead of 1M pre-training-grade (ablation)Llama 3.2 3B + Rad-DINO · Structured VQARadStrucVQA Logical Prec.: 0.67 -> 0.76Improved
add 1M fine-tuning-grade stage after 1M pre-training-grade (ablation)Llama 3.2 3B + Rad-DINO · Structured VQARadStrucVQA Logical Prec.: 0.67 -> 0.78Improved
PT(1M) then FT(1M) instead of 2M pre-training samples (ablation)Llama 3.2 3B + Rad-DINO · Structured VQARadStrucVQA Logical Prec.: 0.68 -> 0.78Improved
double pre-training samples from 1M to 2M (ablation)Llama 3.2 3B + Rad-DINO · Structured VQARadStrucVQA Logical Prec.: 0.67 -> 0.68No clear effect
fine-tune on FT-grade data instead of PT-grade (ablation)Llama 3.2 3B + Rad-DINO · Structured VQAGrounding Rec.: 0.92 -> 0.89Hurt
train on CXR-QBA vs zero-shot MAIRA-2Llama 3.2 3B + Rad-DINO · Structured VQARadStrucVQA Logical Prec.: 0.25 -> 0.78Improved
train on CXR-QBA vs zero-shot Qwen3-VLLlama 3.2 3B + Rad-DINO · Structured VQARadStrucVQA Grounding Rec.: 0.51 -> 0.90Improved
LLM-based scene graph with fine-grained tags vs Chest ImaGenome graphsCXR-QBA scene graphs · CXR-LT 2024 gold long-tail classesMCC: 0.59 -> 0.71Improved
our scene graph boxes vs Chest ImaGenomeCXR-QBA scene graphs · MS-CXRIoU@30: 0.45 -> 0.51Improved
our scene graph boxes vs Chest ImaGenomeCXR-QBA scene graphs · MS-CXRIoT@30: 0.98 -> 0.94Hurt
our scene graph boxes vs Chest ImaGenomeCXR-QBA scene graphs · MS-CXRIoP@30: 0.48 -> 0.56Improved
Llama 3.1 8B judge vs larger judges in RadStrucVQA (ablation)RadStrucVQA metric · entailment sub-metricscorrelation: 0.88 to 0.97No clear effect

Closest papers by meaning

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.

Open scime draft

Get the next scime Dispatch and a monthly field update

Free. We send a confirmation link first; every email has a one-click unsubscribe. Privacy