Tag
Datalab is releasing a new processor that converts PDFs to JATS XML, reducing cost to cents per page and time to under 5 minutes with 92.6% accuracy in a human-matched benchmark.
This paper introduces intra-paper claim verification, a framework that uses LLMs to evaluate whether novelty claims in a paper are supported by its methodological evidence, addressing a gap in existing automated peer review systems. Human evaluation shows significant alignment with human reviewer concerns, especially for novelty-related issues.
A Twitter thread analyzes the unusually long peer review process for a cell embedding paper using GPT-5.6 to compare the preprint and final publication, estimating time, compute, personnel, and APC costs, highlighting the cost-benefit ratio of journal peer review.
Science historians discover that the journal Naturwissenschaften retracted two 1940s papers by Max Planck, likely due to a copyright-related algorithm error, leading to blank pages and empty PDFs.
A new study demonstrates that AI-assisted peer review is vulnerable to low-cost manipulation via superficial rephrasing of paper abstracts, significantly inflating AI-generated review scores and potentially biasing human editorial decisions, highlighting the need for safeguards.
Tim O'Reilly discusses the challenges of integrating AI into scientific publishing, including hallucinated citations, propagation of retracted papers, and training on compromised literature, and calls for adapting existing scientific infrastructure for AI use.