dailyai.report

A new dataset called GoodPoint-ICLR contains 19,000 papers with reviewer feedback annotated via author responses. The researchers developed a training recipe to generate targeted, actionable critiques based on validity and author action. This approach shifts LLM utility from automating research to augmenting human peer review. Practitioners can use these signals to refine academic writing.