Multimodal Large Language Models for Medical Report Generation via Customized Prompt Tuning

Medical report generation from imaging data remains a challenging task in clinical practice. While large language models (LLMs) show great promise in addressing this challenge, their effective integration with medical imaging data still deserves in-depth exploration. In this paper, we present MRG-LLM, a novel multimodal large language model (MLLM) that combines a frozen LLM with a learnable visual encoder and introduces a dynamic prompt customization mechanism. Our key innovation lies in generating instance-specific prompts tailored to individual medical images through conditional affine transformations derived from visual features. We propose two implementations: prompt-wise and promptbook-wise customization, enabling precise and targeted report generation. Extensive experiments on IU X-ray and MIMIC-CXR datasets demonstrate that MRG-LLM achieves state-of-the-art performance in medical report generation. Our code will be made publicly available.

Deep Knowledge Tracing with Convolutions

Knowledge tracing (KT) has recently been an active research area of computational pedagogy. The task is to model students mastery level of knowledge based on their responses to the questions in the past, as well as predict the probabilities that they correctly answer subsequent questions in the future. A good KT model can not only make students timely aware of their knowledge states, but also help teachers develop better personalized teaching plans for students. KT tasks were historically solved using statistical modeling methods such as Bayesian inference and factor analysis, but recent advances in deep learning have led to the successive proposals that leverage deep neural networks, including long short-term memory networks, memory-augmented networks and self-attention networks. While those deep models demonstrate superior performance over the traditional approaches, they all neglect more or less the impact on knowledge states of the most recent questions answered by students. The forgetting curve theory states that human memory retention declines over time, therefore knowledge states should be mostly affected by the recent questions. Based on this observation, we propose a Convolutional Knowledge Tracing (CKT) model in this paper. In addition to modeling the long-term effect of the entire question-answer sequence, CKT also strengthens the short-term effect of recent questions using 3D convolutions, thereby effectively modeling the forgetting curve in the learning process. Extensive experiments show that CKT achieves the new state-of-the-art in predicting students performance compared with existing models. Using CKT, we gain 1.55 and 2.03 improvements in terms of AUC over DKT and DKVMN respectively, on the ASSISTments2009 dataset. And on the ASSISTments2015 dataset, the corresponding improvements are 1.01 and 1.96 respectively.