Abstractive summarization models face significant challenges in specialized domains like medical research. Traditional models have a hard time understanding detailed scientific studies and medical terms. This limitation can lead to the generation of “hallucinated” summaries that include information that is not present in the original texts(Leopold et al. in Energies 17:2492, 2024). Such inaccuracies are particularly problematic in healthcare and scientific research, where the precision of information can save lives. Popular techniques used in abstractive summarization are neural networks and trans formers. In this work, we took a closer look at the transformers technique, which has revolutionized abstractive summarization by capturing long-range dependencies in the text. We applied a transformer-based model called PEGASUS on a PubMed dataset and compared the generated summaries with corresponding reference summaries/abstracts to fine-tune the model further. A key focus of this work is maintaining factual accuracy by implementing validation mechanisms to minimize hallucinated content, such that the generated summaries reflect the original research findings.