MIRAI tries to predict which research will matter years later

Researchers introduced MIRAI, a system designed to predict and generate high-impact academic research. Trained on the arXiv academic graph, it forecasts five-year citation counts and PageRank-style influence for papers, reporting Spearman correlations of 0.6192 for citation prediction and 0.4686 for PageRank prediction on 2021 papers. The project also explores generating research ideas likely to become influential. Predicting citations is not the same as judging scientific truth or social value, and such systems could reinforce existing popularity biases if used carelessly. Still, MIRAI is an interesting attempt to measure whether AI can reason about the future trajectory of science rather than only summarize existing literature. The useful next test is whether its forecasts generalize across fields and identify overlooked work.

Key details: June 3, 2026, arXiv academic graph, Five-year impact prediction, 0.6192 Spearman correlation for citations, 0.4686 for PageRank, Research-idea generation.

Continue swiping for more AI Brief stories.