Research map / Vision-language and multimodal / Multimodal LLMs

Vision-Language Models Guided Graph Concept Reasoning for Interpretable Diabetic Retinopathy Diagnosis

AAAI 2026 · Vision-language and multimodal › Multimodal LLMs · approach: Multimodal LLM benchmarks · 0 citations (OpenAlex)

Read the paperCheck your idea against 108,967 papers

AI-extracted summary. Written by AI from the paper's abstract only. Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.

Proposes a vision-language-model-guided graph concept bottleneck approach for interpretable diabetic retinopathy grading with concept-level reasoning and test-time intervention.

Problem

Deep networks for diabetic retinopathy diagnosis are black boxes, limiting clinical acceptance; existing concept bottleneck DR methods model lesions as concepts but (per the truncated abstract) appear to ignore relations among concepts.

Approach

Uses vision-language models to guide a graph-based concept reasoning layer over lesion concepts within a concept bottleneck framework, so grading is driven by relational concept structure rather than independent lesion concepts. Details beyond the abstract are not available.

Experimental setup

Builds on: concept bottleneck model

Closest papers by meaning

• FOCUS: Benchmarking Retinal Model Generalization from Foundation Vision Encoders to Multimodal LLMs (NeurIPS 2026 · similarity 0.87)
• HACDR-Net: Heterogeneous-Aware Convolutional Network for Diabetic Retinopathy Multi-Lesion Segmentation (AAAI 2024 · similarity 0.87)
• DrVD-Bench: Do Vision-Language Models Reason Like Human Doctors in Medical Image Diagnosis? (NeurIPS 2025 · similarity 0.86)
• Like an Ophthalmologist: Dynamic Selection Driven Multi-View Learning for Diabetic Retinopathy Grading (AAAI 2025 · similarity 0.86)
• Visual and Domain Knowledge for Professional-level Graph-of-Thought Medical Reasoning (ICML 2025 · similarity 0.85)
• SIGraph: Saliency Image-Graph Network for Retinal Disease Classification in Fundus Image (AAAI 2025 · similarity 0.85)