"We were shocked when we saw the Pangram results,"
"was disconcerting to see people increasing the usage of LLMs for peer review in spite of us prohibiting that usage"
"our intention is definitely to start screening all incoming manuscripts and all incoming peer review comments"
"seems to work exceptionally well"
Presence of AI-generated text in research submissions and peer-review reports rose sharply after late 2022. In 2024, 23% of abstracts and 5% of peer-review reports contained text probably generated by large language models. Less than 25% of authors disclosed AI use despite mandatory disclosure requirements. Screening covered 46,500 abstracts, 46,021 methods sections, and 29,544 peer-review comments submitted to 10 journals between 2021 and 2024. Detections declined by 50% in late 2023 after a ban on LLM use for peer reviewers, then more than doubled by early 2024. Publishers intend to screen all incoming manuscripts and peer-review comments.
Read at Nature
Unable to calculate read time
Collection
[
|
...
]