Cross-topic name collisions grow 4.33-fold in AI papers
Original title: "PRISM" Is Not All You Need: Cross-Topic Name Collisions and Scientific Paper Retrieval
Shared names create 25.7 percentage point identification deficits in retrieval for cross-topic cases.
Scientific names increasingly link papers across research topics. Study of 564,861 AI-related arXiv records shows cross-topic collisions rise 4.33-fold from 2020-2022 reference to 2026. A second sentence encoder and name-group resampling confirm the pattern. Context resolves more ambiguity than title or abstract alone.
- Collisions grow 4.33-fold cross-topic
- 25.7 point deficit with BM25
- 6.9 points with E5
- Listwise LM ranking reaches 89.2%
- Intermediate-frequency words drive increase
From the original article
1 min read at huggingface.co
Scientific names increasingly connect papers from different research topics, leaving readers to recover a paper's identity from context. We study this convergence in 564,861 AI-related arXiv records. Cross-topic name collision grows 4.
By Jingjie Ning, Xueqi Li, Yibo Kong · huggingface.co
A short excerpt; the article and its images belong to the publisher. Summary, key facts and assessment above are AI-written from the article and can be wrong.