Prompt tuning is one of the most effective solutions to adapting a fixed pre-trained language model (PLM) for various downstream tasks, especially with only a few input samples. However, the security issues, e.g., Trojan attacks, of prompt tuning on a few data samples are not well-studied. Transferring established data poisoning attacks directly to few-shot prompt tuning presents multiple challenges. One significant issue is the _poisoned imbalance issue_, where non-target class samples are added to the target class, resulting in a greater number of target-class samples compared to non-target class. While this issue is not critical in regular tuning, it significantly hampers the few-shot prompt tuning, making it difficult to simultaneously achieve a high attack success rate (ASR) and maintain clean data accuracy (CDA). Additionally, few-shot prompting is prone to overfitting in terms of both ASR and CDA. In this paper, we introduce _TrojFSP_, a method designed to address the challenges. To solve the poisoned imbalance issue, we develop a _Target-Class Shrink (TC-Shrink)_ technique, which aims to equalize the number of poisoning samples. To combat overfitting, we employ a _Selective Token Poisoning_ technique to boost attack performance. Furthermore, we introduce a _Trojan-Trigger Attention_ objective function to amplify the attention of the poisoned trojan prompt on triggers. Experiments show that our TrojFSP achieves an ASR of over 99% while maintaining negligible decreases in CDA across various PLMs and datasets. The source code of TrojFSP is available at _https://github.com/UCF-ML-Research/TrojFSP_.

This paper introduces the RAG-RLRC-LaySum framework, designed to make complex biomedical research accessible to laymen through advanced Natural Language Processing (NLP) techniques. Our innovative Retrieval Augmentation Generation (RAG) solution, enhanced by a reranking method, utilizes multiple knowledge sources to ensure the precision and pertinence of lay summaries. Additionally, our Reinforcement Learning for Readability Control (RLRC) strategy improves readability, making scientific content comprehensible to non-specialists. Evaluations using the publicly accessible PLOS and eLife datasets show that our methods surpass Plain Gemini model, demonstrating a 20% increase in readability scores, a 15% improvement in ROUGE-2 relevance scores, and a 10% enhancement in factual accuracy. The RAG-RLRC-LaySum framework effectively democratizes scientific knowledge, enhancing public engagement with biomedical discoveries.

A human decision-maker benefits the most from an AI assistant that corrects for their biases. For problems such as generating interpretation of a radiology report given findings, a system predicting only highly likely outcomes may be less useful, where such outcomes are already obvious to the user. To alleviate biases in human decision-making, it is worth considering a broad differential diagnosis, going beyond the most likely options. We introduce a new task, “less likely brainstorming,” that asks a model to generate outputs that humans think are relevant but less likely to happen. We explore the task in two settings: a brain MRI interpretation generation setting and an everyday commonsense reasoning setting. We found that a baseline approach of training with less likely hypotheses as targets generates outputs that humans evaluate as either likely or irrelevant nearly half of the time; standard MLE training is not effective. To tackle this problem, we propose a controlled text generation method that uses a novel contrastive learning strategy to encourage models to differentiate between generating likely and less likely outputs according to humans. We compare our method with several state-of-the-art controlled text generation models via automatic and human evaluations and show that our models’ capability of generating less likely outputs is improved.

Generating a summary from findings has been recently explored (Zhang et al., 2018, 2020) in note types such as radiology reports that typically have short length. In this work, we focus on echocardiogram notes that is longer and more complex compared to previous note types. We formally define the task of echocardiography conclusion generation (EchoGen) as generating a conclusion given the findings section, with emphasis on key cardiac findings. To promote the development of EchoGen methods, we present a new benchmark, which consists of two datasets collected from two hospitals. We further compare both standard and start-of-the-art methods on this new benchmark, with an emphasis on factual consistency. To accomplish this, we develop a tool to automatically extract concept-attribute tuples from the text. We then propose an evaluation metric, FactComp, to compare concept-attribute tuples between the human reference and generated conclusions. Both automatic and human evaluations show that there is still a significant gap between human-written and machine-generated conclusions on echo reports in terms of factuality and overall quality.

Rapid growth in adoption of electronic health records (EHRs) has led to an unprecedented expansion in the availability of large longitudinal datasets. Large initiatives such as the Electronic Medical Records and Genomics (eMERGE) Network, the Patient-Centered Outcomes Research Network (PCORNet), and the Observational Health Data Science and Informatics (OHDSI) consortium, have been established and have reported successful applications of secondary use of EHRs in clinical research and practice. In these applications, natural language processing (NLP) technologies have played a crucial role as much of detailed patient information in EHRs is embedded in narrative clinical documents. Meanwhile, a number of clinical NLP systems, such as MedLEE, MetaMap/MetaMap Lite, cTAKES, and MedTagger have been developed and utilized to extract useful information from diverse types of clinical text, such as clinical notes, radiology reports, and pathology reports. Success stories in applying these tools have been reported widely. Despite the demonstrated success of NLP in the clinical domain, methodologies and tools developed for the clinical NLP are still underknown and underutilized by students and experts in the general NLP domain, mainly due to the limited exposure to EHR data. Through this tutorial, we would like to introduce NLP methodologies and tools developed in the clinical domain, and showcase the real-world NLP applications in clinical research and practice at Mayo Clinic (the No. 1 national hospital ranked by the U.S. News & World Report) and the University of Minnesota (the No. 41 best global universities ranked by the U.S. News & World Report). We will review NLP techniques in solving clinical problems and facilitating clinical research, the state-of-the art clinical NLP tools, and share collaboration experience with clinicians, as well as publicly available EHR data and medical resources, and finally conclude the tutorial with vast opportunities and challenges of clinical NLP. The tutorial will provide an overview of clinical backgrounds, and does not presume knowledge in medicine or health care. The goal of this tutorial is to encourage NLP researchers in the general domain (as opposed to the specialized clinical domain) to contribute to this burgeoning area. In this tutorial, we will first present an overview of clinical NLP. We will then dive into two subareas of clinical NLP in clinical research, including big data infrastructure for large-scale clinical NLP and advances of NLP in clinical research, and two subareas in clinical practice, including clinical information extraction and patient cohort retrieval using EHRs. Around 70% of the tutorial will review clinical problems, cutting-edge methodologies, and real-world clinical NLP tools while another 30% introduce use cases at Mayo Clinic and the University of Minnesota. Finally, we will conclude the tutorial with challenges and opportunities in this rapidly developing domain.