Context
Original research on a 250-paper computer-science benchmark finds that breadth-first bibliography expansion greatly improves scholarly retrieval recall and that reference-list overlap alone does not capture topical citation relevance.
What changed
A full-document, multi-query scholarly retrieval pipeline combines API seeds, bibliography expansion and reranking; the authors separately measure recall against human references, LLM-judged relevance, diversity and co-authorship proximity.
Why it matters
AI discovery measurement should separate whether sources were found from how they were ranked and judged relevant; existing human citations are not a complete or unambiguous ground truth for evaluating retrieval.
Visibility OS interpretation
Visibility OS interpretation: this controlled scholarly-search result suggests measuring retrieval coverage, citation relevance and evidence diversity separately. It is not a test of live commercial answer-engine citation ranking.
What remains uncertain
Will independent human judges reproduce the LLM-judged topical relevance gap? Do findings generalize beyond arXiv computer-science papers and beyond scholarly search? How much of recall improvement is due to larger candidate budgets?