Paper acceptance predictors fail to identify beneficial revisions

A new paper published on arXiv investigates the effectiveness of paper acceptance predictors, specifically focusing on their ability to identify improvements in submitted manuscripts. Researchers compared a classifier using word and document features with two existing text-only predictors fine-tuned from a large language model. The study found that while the predictors could rank original papers similarly, they showed no significant agreement on whether paragraph rewrites improved a paper's score. This lack of agreement persisted even after re-scoring with higher precision, suggesting that current predictors may not reliably identify beneficial revisions. AI

IMPACT Suggests current AI models used for paper review may not reliably identify substantive improvements, impacting academic publishing workflows.

RANK_REASON The cluster contains a research paper published on arXiv detailing findings about the performance of AI models. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.CL →

• alphaXiv
• arXiv
• CatalyzeX Code Finder for Papers
• Connected Papers
• CORE Recommender
• DagsHub
• Gotit.pub
• Hugging Face
• Influence Flower
• International Conference on Learning Representations
• Litmaps
• ScienceCast
• scite Smart Citations

• paper
• other

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →