Research map / Vision-language and multimodal / Cross-modal learning
DiA-gnostic VLVAE: Disentangled Alignment-Constrained Vision Language Variational AutoEncoder for Robust Radiology Reporting with Missing Modalities
Nagur Shareef Shaik, Teja Krishna Cherukuri, Adnan Masood, Dong Hye Ye
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 VL-MoE-VAE to baseline (with context) (ablation) | DiA baseline · MIMIC-CXR (with context) | clinical F1: 0.446 -> 0.551 | Improved |
| add VL-MoE-VAE to baseline (with context) (ablation) | DiA baseline · IU X-Ray (with context) | BLEU-4: 0.262 -> 0.319 | Improved |
| add disentangled alignment (orthogonality + contrastive) constraint (ablation) | Baseline + VL-MoE-VAE · MIMIC-CXR (with context) | clinical F1: 0.551 -> 0.621 | Improved |
| add disentangled alignment constraint (ablation) | Baseline + VL-MoE-VAE · MIMIC-CXR (missing context) | clinical F1: 0.394 -> 0.438 | Improved |
| add VL-MoE-VAE and DA when clinical context is missing (ablation) | DiA baseline · IU X-Ray (missing context) | BLEU-4: 0.079 -> 0.198 | Improved |
| use custom end-to-end encoder instead of RAD-DINO + CXR-BERT (ablation) | DiA · MIMIC-CXR | clinical F1: 0.441 -> 0.497 | Improved |
| use LLaMA-X decoder instead of standard Transformer decoder (ablation) | DiA · MIMIC-CXR | clinical F1: 0.479 -> 0.497 | Improved |
| use LLaMA-X decoder instead of GPT-2 decoder (ablation) | DiA · MIMIC-CXR | clinical F1: 0.419 -> 0.497 | Improved |
| DiA vs SEI on MIMIC-CXR BLEU-4 | DiA · MIMIC-CXR | BLEU-4: 0.135 (SEI) vs 0.134 (DiA) | No clear effect |
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.
- PromptMRG — PromptMRG: Diagnosis-Driven Prompts for Medical Report Generation (AAAI 2024) · better on 1 of 1 shared result
- KiUT — KiUT: Knowledge-Injected U-Transformer for Radiology Report Generation (CVPR 2023) · 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
- RadZero: Similarity-Based Cross-Attention for Explainable Vision-Language Alignment in Chest X-ray with Zero-Shot Multi-Task Capability (NeurIPS 2025 · similarity 0.86)
- S2D-Align: Shallow-to-Deep Auxiliary Learning for Anatomically-Grounded Radiology Report Generation (AAAI 2026 · similarity 0.86)
- Toward a Vision-Language Foundation Model for Medical Data: Multimodal Dataset and Benchmarks for Vietnamese PET/CT Report Generation (NeurIPS 2025 · similarity 0.85)
- Large-scale and Fine-grained Vision-language Pre-training for Enhanced CT Image Understanding (ICLR 2025 · similarity 0.85)
- PairAug: What Can Augmented Image-Text Pairs Do for Radiology? (CVPR 2024 · similarity 0.85)
- RNED: Rotary Number Encoding and Decoding for Medical VLMs (CVPR 2026 · similarity 0.85)
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.
