A study published in The BMJ used a BERT-based “scientific spam filter,” trained on 2,202 retracted paper-mill papers, to screen 2.6 million cancer studies from 1999 to 2024. Led by Queensland University of Technology biostatistician Adrian Barnett, the analysis flagged 261,245 papers, or 9.87%, with suspicious writing patterns, rising from about 1% in the early […]
