Research map / Vision-language and multimodal / Visual question answering
MedFG-VQA: Low-Frequency Memory and Graph Attention for Lightweight Medical VQA
Haowen Gu, Gensheng Pei, Zeren Sun, Mingwu Ren, Xiangbo Shu, Yazhou Yao, Fumin Shen
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 FMF and GACA modules (ablation) | MedFG-VQA baseline · SynMedVQA average | accuracy: 0.627 -> 0.6441 | Improved |
| use DCT instead of FFT in FMF (ablation) | FMF module · SynMedVQA average | accuracy: 0.5585 -> 0.6441 | Improved |
| add DCT frequency transform versus none (ablation) | FMF module · SynMedVQA average | accuracy: 0.6058 -> 0.6441 | Improved |
| use FFT instead of no transform (ablation) | FMF module · SynMedVQA average | accuracy: 0.6058 -> 0.5585 | Hurt |
| replace standard cross-attention with GACA (ablation) | cross-modal attention · SynMedVQA average | accuracy: 0.6407 -> 0.6441 | Improved |
| increase memory bank size from 16 to 64 (ablation) | FMF memory · SynMedVQA average | accuracy | Improved |
| increase memory bank size from 64 to 128 (ablation) | FMF memory · SynMedVQA average | accuracy | Hurt |
| vary diversity loss weight lambda from 0.3 to 0.9 (ablation) | FMF diversity loss · SynMedVQA average | accuracy: 0.6441 | No clear effect |
| MedFG-VQA vs Qwen3-VL on SynMedVQA | Qwen3-VL · SynMedVQA average | accuracy: 0.5492 -> 0.6441 | Improved |
| MedFG-VQA vs Qwen3-VL on closed-ended VQA-RAD | Qwen3-VL · VQA-RAD closed | accuracy: 0.7868 -> 0.6324 | Hurt |
| MedFG-VQA vs Qwen3-VL on open-ended PathVQA | Qwen3-VL · PathVQA open | accuracy: 0.5249 -> 0.8062 | Improved |
Closest papers by meaning
- Alignment, Mining and Fusion: Representation Alignment with Hard Negative Mining and Selective Knowledge Fusion for Medical Visual Question Answering (CVPR 2025 · similarity 0.92)
- Falcon Medical Visual Question Answering (AAAI 2025 · similarity 0.91)
- DiN: Diffusion Model for Robust Medical VQA with Semantic Noisy Labels (CVPR 2025 · similarity 0.90)
- OmniMedVQA: A New Large-Scale Comprehensive Evaluation Benchmark for Medical LVLM (CVPR 2024 · similarity 0.90)
- Fine-grained Adaptive Visual Prompt for Generative Medical Visual Question Answering (AAAI 2025 · similarity 0.90)
- Visual and Domain Knowledge for Professional-level Graph-of-Thought Medical Reasoning (ICML 2025 · similarity 0.90)
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.
