Paper-reported evidence

ADAR-GPT: A continually fine-tuned language model for predicting A-to-I RNA editing sites

2026 · Peer-reviewed · version of record

Paper-reported results. The values below come from this source, not an independent rewire.it run.

Open primary paper →DOI: 10.1073/pnas.2529073123

Primary full text verified via Europe PMC fullTextXML; venue: Proceedings of the National Academy of Sciences of the United States of America; PMC ID: PMC12798952. RNA editing site benchmark on a restricted liver validation set.

1 verified numerical rows · source checked 2026-09-15

Paper-reportedAuthor's model

ADAR-GPT continual

A-to-I RNA editing site prediction

Reported score: 0.763
Metric: F1
Dataset / split: liver editing sites · 15% validation set

Curriculum plus 15% fine-tuning; 201-nt sequence windows; decision threshold 0.5

Table 2, Adar-GPT (continual) row, F1 columnVerify at source →