Research map / Vision-language and multimodal / Multimodal LLMs
Uncertainty-Driven Expert Control: Enhancing the Reliability of Medical Vision-Language Models
Xiao Liang, Di Wang, Zhicheng Jiao, Ronghan Li, Pengfei Yang, Quan Wang, Tat-Seng Chua
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 |
|---|---|---|---|
| apply Expert-CFG to top 5% high-entropy predictions | Phi3.5V-Med 4.2B · VQA-RAD overall | accuracy/recall: 82.04 -> 84.70 | Improved |
| apply Expert-CFG to top 5% high-entropy predictions | Phi3V-Med 4.2B · PathVQA overall | accuracy/recall: 64.43 -> 68.91 | Improved |
| use classifier-free guidance instead of plain expert-annotated RAG (ablation) | Phi3.5V-Med 4.2B (5% entropy subset) · VQA-RAD overall | accuracy/recall: 83.37 -> 84.70 | Improved |
| retrieval-augmented generation on all queries (ablation) | Phi3.5V-Med 4.2B · VQA-RAD overall | accuracy/recall: 82.04 -> 81.15 | No clear effect |
| retrieval-augmented generation on all queries (ablation) | Phi3.5V-Med 4.2B · PathVQA overall | accuracy/recall: 65.24 -> 72.90 | Improved |
| restrict expert annotations to top 5% highest-entropy samples (ablation) | Expert-CFG · three MedVQA benchmarks | accuracy | Improved |
| use predictive entropy to detect unreliable answers | Phi3.5V-Med 4.2B · VQA-RAD/SLAKE/PathVQA | AUC: AUC exceeds 0.8 across all datasets | Improved |
| set alpha to 0 (fully mask highlighted tokens in unconditional branch) (ablation) | Expert-CFG · VQA-RAD overall (100% annotation) | accuracy: 98.44 -> 96.23 | Hurt |
| raise guidance strength gamma from 1.3 to 1.5 (ablation) | Expert-CFG · PathVQA overall (100% annotation) | accuracy: 94.76 -> 92.03 | Hurt |
| retrieve with union of separate image and text features instead of image only (ablation) | BioMedCLIP retrieval · SLAKE open-ended | top-1 hit rate: 47.75 -> 55.97 | Improved |
| train Phi3.5V on PubMedVision instead of no medical tuning (ablation) | Phi3.5V 4.2B · zero-shot closed QA VQA-RAD | accuracy: 57.72 -> 74.63 | Improved |
| use PubMedVision instead of LLaVA-Med data (ablation) | Phi3V 4.2B · zero-shot closed QA SLAKE | accuracy: 60.58 -> 67.31 | Improved |
Closest papers by meaning
- VILA-M3: Enhancing Vision-Language Models with Medical Expert Knowledge (CVPR 2025 · similarity 0.91)
- Towards a Multimodal Large Language Model with Pixel-Level Insight for Biomedicine (AAAI 2025 · similarity 0.90)
- MediConfusion: Can you trust your AI radiologist? Probing the reliability of multimodal medical foundation models (ICLR 2025 · similarity 0.89)
- OpenMedReason: Scientific Reasoning Supervision for Medical Vision–Language Models (NeurIPS 2026 · similarity 0.89)
- OmniMedVQA: A New Large-Scale Comprehensive Evaluation Benchmark for Medical LVLM (CVPR 2024 · similarity 0.88)
- Alignment, Mining and Fusion: Representation Alignment with Hard Negative Mining and Selective Knowledge Fusion for Medical Visual Question Answering (CVPR 2025 · similarity 0.88)
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.
