Finding a paper from an informal research idea requires more than matching topic words: relevant work may use different terminology, and its distinguishing mechanism may be absent from the abstract. We present \system{}, an auditable literature-search system built on PaperArena. It combines multi-source discovery, identity verification, facet-based abstract assessment, semantic citation expansion, and joint Expert ranking. A retrieval-only mode ranks up to five papers from a shortlist of at most 60, using complete available abstracts and locally selected PDF evidence from five fixed acquisition attempts, without generating a long-form report. In a historical comparison covering ten queries from each of EASY, MEDIUM, and HARD, aggregate reported Hit@5 is 53.3\% for \system{}, 43.3\% for ChatGPT web Deep Research, and 36.7\% for Gemini web Deep Research under the documented accounting convention. The advantage is concentrated in EASY and MEDIUM; ChatGPT leads on HARD. Selected \system{} runs average 8 minutes 53 seconds, compared with user-reported web averages of 5.2 and 4.7 minutes. A separate nine-query API sample averages \$0.0945 per task; this is a historical cost observation, not a matched comparison with web subscription costs. Pairwise Top 5 coverage ranges from 6\% to 34\%, showing differing returned evidence sets. The system also exposes optional implementation analysis and citation-evolution tools. These results support targeted idea-to-paper retrieval while leaving general web research, forum search, and broader relevance evaluation open. https://github.com/ystok66/IDEA2PRIOR

翻译：暂无翻译