GPT-4’s purported strong multimodal abilities raise interests in using it to automate radiology report writing, but there lacks thorough evaluations. In this work, we perform a systematic evaluation of GPT-4 (4o and vision-preview) in generating radiology reports across three chest X-ray report benchmarks: MIMIC-CXR, CheXpert Plus, and IU X-Ray. We attempt to directly generate reports with different prompting strategies and find that the models fail terribly in both lexical metrics and clinical efficacy metrics. To understand the low performance, we decompose the task into two steps: 1) the **medical image reasoning** step of predicting medical condition labels from images; and 2) the **report synthesis** step of generating reports from (groundtruth) conditions. We show that GPT-4’s performance in image reasoning is consistently low across different prompts. In fact, the distributions of model-predicted labels remain constant regardless of which groundtruth conditions are present on the image, suggesting that the model is not interpreting chest X-rays meaningfully. Even when given groundtruth conditions in report synthesis, its generated reports are less correct and less natural-sounding than a finetuned Llama. Altogether, our findings cast doubt on the viability of using GPT-4 in a radiology workflow.

Large language models (LLMs) still lack delicate controllability over their responses, which is critical to enhancing their performance and the user experience. However, curating supervised fine-tuning (SFT) datasets to improve LLM controllability usually relies on human experts or proprietary LLMs, which requires additional costs. To bridge this gap, we propose Rule-based Data Recycling (RuleR), a data augmentation method incorporating multiple constraints into the original data samples according to predefined rules, which creates new training tasks to consolidate the controllability of LLMs. Instead of creating new data from scratch, RuleR “recycles” existing data by simply applying rule-based edits to their responses and appending the rule-instructions in their original instructions. Experimental results demonstrate RuleR’s effectiveness in improving LLM controllability while maintaining general instruction-following capabilities.

Large language models (LLMs), despite their breakthroughs on many challenging benchmark tasks, prefer to generate verbose responses and lack the controllability of output complexity, which is usually preferred by human users in practice. In this paper, we study how to precisely control multiple linguistic complexities of LLM output by finetuning using off-the-shelf data. To this end, we propose multi-control tuning (MCTune), which includes multiple linguistic complexity values of ground-truth responses as controls in the input for instruction tuning. We finetune LLaMA2-7B on Alpaca-GPT4 and WizardLM datasets. Evaluations on widely used benchmarks demonstrate that our method does not only improve LLMs’ multi-complexity controllability substantially but also retains or even enhances the quality of the responses as a side benefit.

In this paper, we describe a system that we built to participate in the SemEval 2022 Task 11: MultiCoNER Multilingual Complex Named Entity Recognition, specifically the track Mono-lingual in English. To construct this system, we used Pre-trained Language Models (PLMs). Especially, the Pre-trained Model base on BERT is applied for the task of recognizing named entities by fine-tuning method. We performed the evaluation on two test datasets of the shared task: the Practice Phase and the Evaluation Phase of the competition.

This paper describes the improvements that we have applied on CAMR baseline parser (Wang et al., 2016) at Task 8 of SemEval-2016. Our objective is to increase the performance of CAMR when parsing sentences from scientific articles, especially articles of biology domain more accurately. To achieve this goal, we built two wrapper layers for CAMR. The first layer, which covers the input data, will normalize, add necessary information to the input sentences to make the input dependency parser and the aligner better handle reference citations, scientific figures, formulas, etc. The second layer, which covers the output data, will modify and standardize output data based on a list of scientific concept fixing patterns. This will help CAMR better handle biological concepts which are not in the training dataset. Finally, after applying our approach, CAMR has scored 0.65 F-score on the test set of Biomedical training data and 0.61 F-score on the official blind test dataset.