MedExpMem: Adapting Experience Memory for Differential Diagnosis
The paper introduces MedExpMem, a framework designed to enhance differential diagnosis in medical vision-language models. This framework allows diagnostic agents to accumulate and utilize experience memory derived from their own diagnostic failures. Evaluation results indicate that MedExpMem significantly improves accuracy across various models and scales in radiology benchmarks.
- ▪MedExpMem enables VLM-based diagnostic agents to accumulate differential diagnosis expertise.
- ▪The framework organizes experience as pairwise differential notes, encoding key discriminators and decision rules.
- ▪Results show accuracy improvements of up to 7.0% across diverse models in radiology benchmarks.
Opening excerpt (first ~120 words) tap to expand
Computer Science > Machine Learning arXiv:2605.22872 (cs) [Submitted on 20 May 2026] Title:MedExpMem: Adapting Experience Memory for Differential Diagnosis Authors:Qianhan Feng, Zhongzhen Huang, Yakun Zhu, Yannian Gu, Winnie Chiu Wing Chu, Xiaofan Zhang, Qi Dou View a PDF of the paper titled MedExpMem: Adapting Experience Memory for Differential Diagnosis, by Qianhan Feng and 6 other authors View PDF HTML (experimental) Abstract:Experienced physicians develop diagnostic expertise through clinical practice, acquiring not only disease knowledge but also the ability to differentiate confusable conditions. Current medical vision-language models (VLMs) lack this capability -- their parameters encode static knowledge that does not evolve across diagnostic encounters.
…
Excerpt limited to ~120 words for fair-use compliance. The full article is at arXiv cs.AI.
Discussion
0 commentsMore from arXiv cs.AI
