Human Knowledge Integrated Multi-modal Learning for Single Source Domain Generalization
About
Generalizing image classification across domains remains challenging in critical tasks such as fundus image-based diabetic retinopathy (DR) grading and resting-state fMRI seizure onset zone (SOZ) detection. When domains differ in unknown causal factors, achieving cross-domain generalization is difficult, and there is no established methodology to objectively assess such differences without direct metadata or protocol-level information from data collectors, which is typically inaccessible. We first introduce domain conformal bounds (DCB), a theoretical framework to evaluate whether domains diverge in unknown causal factors. Building on this, we propose GenEval, a multimodal Vision Language Models (VLM) approach that combines foundational models (e.g., MedGemma-4B) with human knowledge via Low-Rank Adaptation (LoRA) to bridge causal gaps and enhance single-source domain generalization (SDG). Across eight DR and two SOZ datasets, GenEval achieves superior SDG performance, with average accuracy of 69.2% (DR) and 81% (SOZ), outperforming the strongest baselines by 9.4% and 1.8%, respectively.
Related benchmarks
| Task | Dataset | Result | Rank | |
|---|---|---|---|---|
| Diabetic Retinopathy (DR) grading | APTOS | Accuracy73.2 | 25 | |
| Diabetic Retinopathy (DR) grading | FGADR | Accuracy56.9 | 20 | |
| Diabetic Retinopathy (DR) grading | IDRID | Accuracy70.6 | 20 | |
| Diabetic Retinopathy Grading | Aptos (held-out) | Accuracy73.46 | 11 | |
| Diabetic Retinopathy Grading | Messidor 2 (held-out) | Accuracy79.64 | 11 | |
| Diabetic Retinopathy Grading | Messidor (held-out) | Accuracy67.7 | 11 | |
| Diabetic Retinopathy Classification | EyePACS | Accuracy80.04 | 6 | |
| Diabetic Retinopathy Classification | Messidor | Accuracy69.48 | 6 | |
| Diabetic Retinopathy Grading | EyePACS (held-out target) | Accuracy83.18 | 6 | |
| Diabetic Retinopathy Classification | APTOS | Accuracy73.16 | 6 |