Revolutionizing Medical Literature: How Language Models are Changing Systematic Reviews

Scientists have been using systematic reviews (SR) to gather and analyze medical research. But checking through all the articles is hard work. Some people thought using language models could help make this easier. These models can pick out which articles are relevant, but they are still being tested to see if they work as well as humans. Researchers tested 18 of these models on three different systematic reviews. They found that these models could correctly choose between 4, 130 and 4, 329 out of 4, 662 titles and abstracts. That's quite a lot! However, each model had different success rates, and some had trouble with certain types of articles. Also, it turned out that how well the models did was affected by how clearly the rules for including or excluding articles were defined. This means that using these models could save reviewers a lot of time, but only if the rules are set up just right first.