[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86561-en":3,"doc-seo-86561-105":30,"detail-sidebar-cat-0-en-105":92},{"code":4,"msg":5,"data":6},0,"success",{"doc_id":7,"user_id":8,"nickname":9,"user_avatar":10,"doc_module":4,"category_id":11,"category_name":12,"doc_title":13,"doc_description":14,"doc_content":15,"file_id":16,"file_url":17,"file_type":18,"file_size":19,"view_count":20,"is_deleted":4,"is_public":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86561,1374391974585,"Genevieve","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","Boolean queries are all you need?","Using an LLM-based search agent connected to a Boolean retrieval engine, this work searches the MS MARCO V2.1 deduped segment collection behind the TREC 2024 RAG track. On a standard subset of 86 topics and under a budget of 100 model calls per topic, the agent reaches NDCG@10 of 0.6863, outperforming many dense, sparse, and learned-sparse first-stage retrievers. Scoring relies only on the density and length of corpus substrings matching a query, without supervised learning, global statistics, or term weights. Results suggest simple pattern matching may suffice for agentic search.","arXiv :2607 . 1 1362v 1 [ cs .IR] 13 Jul 2026  \nBoolean queries are all you need?  \nCharles L. A. Clarke Mark D. Smucker  \nUniversity of Waterloo, Canada  \nAbstract  \nWe equipped an LLM-based search agent with access to a Boolean retrieval engine to search the MS MARCO V2.1 deduped segment collection used by the TREC 2024 RAG track. Over a standard track subset of 86 topics—and operating under a budget of  \n100 model calls/topic—the agent achieved an NDCG@10 of 0.6863, which would place it above many dense, sparse, and learned-sparse first-stage retrievers. Ranking is based solely on the density of corpus substrings matching a query—with no requirement for supervised learning, global statistics, or term weights. Formally, the query language expresses a strict subset of the regular languages, with a document’s score based on the number and length of matches it contains. Although the results are more exploratory than definitive—because they are based on a single test collection that was publicly available during model training—they suggest that simple pattern matching may be sufficient for agentic search.  \n1 Introduction  \nLong before the emergence of AI-powered natural-language search and keyword-focused Web search, Boolean queries provided the primary interface to many operational search engines [24] . While Boolean retrieval was effective in the hands of librarians and other trained technicians—allowing them to specify precise requirements for retrieved results—it proved difficult for untrained people to successfully formulate queries. As early as 1982, efforts were already underway to automatically reformulate natural language queries into Boolean queries, simplifying interaction [23] . By the mid-1990s, Boolean retrieval was fading away. Keyword-based sparse vector retrieval was well established in academic contexts, BM25 was gaining a reputation for excellent effectiveness, and the earliest keyword-based Web search engines were widely available [22, 25, 34] . Even so, as late as 1995, manually constructed Boolean queries could still outperform the best automatic systems at TREC [7, 14] . A few years later, using interactive searching and judging with Boolean queries, trained searchers independently created judgments yielding system rankings consistent with official TREC-6 rankings, translating human intuition into exact-match queries to find relevant documents [10, 31] . Nonetheless, given that untrained humans often struggle to formulate effective Boolean queries, by the 21st century, they were largely abandoned for general-purpose search.  \n(^ \"goth rock\" (+ society social political issues themes addressed))  \n(^ \"goth rock\"  \n(+ darkness depravity alienation despair death)  \n(+ lyrics themes))(^ goth  \n\"post-punk\"  \n(+ politics political \"social commentary\" society)  \n(+ lyrics themes))  \nFigure 1: Example LLM-generated queries for the TREC 2024 RAG Track topic “what society issues did goth rock address” (\\#2024-152259) .  \n(^  \n(+  \n(^ quebec (+ separatism separatist separation separate))  \n(^ quebec (+ sovereignty sovereign independent independence))  \n\"jacques parizeau\"  \n\"louise beaudoin\"  \n\"parti quebecois\")  \n(+ economic economically economics poll opinion referendum majority))  \nFigure 2: Example human-generated query for the TREC-4 adhoc topic “What are the prospects of the Quebec separatists achieving independence from the rest of Canada?”(\\#207), taken from the query set of Clarke et al. [7] with notation adjusted for consistency with the modern Cottontail notation of Figure 1 .  \nLarge language models face no such struggle. When appropriately prompted, LLMs can reformulate natural language into reasonable-seeming Boolean queries [1, 27, 32, 33] . Figure 1 provides an example from the experiments in this paper. This example can be compared to the human-generated query in Figure 2 . In both cases, the Boolean query includes terms not found in the original topics. In both cases, the searcher was encouraged to u","cbCaieReJSRzoeSj","https://ap.wps.com/l/cbCaieReJSRzoeSj","pdf",224402,6,1,11,"English","en",105,"# Introduction\n## Background on Boolean retrieval\n## Motivation for LLM-generated Boolean queries\n# Method\n## Vole search agent and interaction budget\n## Boolean string-matching and re-ranking","[{\"question\":\"How does the paper’s search agent work?\",\"answer\":\"It uses an LLM as a controller that iteratively issues Boolean queries to a Boolean retrieval engine and then requests snippets or full documents. The agent collects relevance judgments with the LLM and stops when it reaches the call/judgment limits.\"},{\"question\":\"What determines document ranking in the Boolean retrieval setup?\",\"answer\":\"Documents are ranked purely by the length and number of shortest substring matches to the Boolean query expressions. The method avoids supervised learning, global statistics, and term weights.\"},{\"question\":\"What performance result does the agent achieve on the TREC 2024 RAG track subset?\",\"answer\":\"On 86 topics with a budget of 100 model calls per topic, the agent achieves NDCG@10 of 0.6863, placing it above many dense, sparse, and learned-sparse first-stage retrievers.\"}]",1784212636,28,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":87,"head_meta":89,"extra_data":91,"updated_unix":28},"boolean-queries-are-all-you-need","",{"@graph":36,"@context":86},[37,54,69],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":53},"https://docshare.wps.com/document/boolean-queries-are-all-you-need/86561/",4,{"url":52,"name":13,"@type":55,"author":56,"headline":13,"publisher":58,"fileFormat":61,"inLanguage":24,"description":14,"dateModified":62,"datePublished":63,"encodingFormat":61,"isAccessibleForFree":64,"interactionStatistic":65},"DigitalDocument",{"name":9,"@type":57},"Person",{"url":41,"name":59,"@type":60},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":66,"interactionType":67,"userInteractionCount":20},"InteractionCounter",{"@type":68},"ViewAction",{"@type":70,"mainEntity":71},"FAQPage",[72,78,82],{"name":73,"@type":74,"acceptedAnswer":75},"How does the paper’s search agent work?","Question",{"text":76,"@type":77},"It uses an LLM as a controller that iteratively issues Boolean queries to a Boolean retrieval engine and then requests snippets or full documents. The agent collects relevance judgments with the LLM and stops when it reaches the call/judgment limits.","Answer",{"name":79,"@type":74,"acceptedAnswer":80},"What determines document ranking in the Boolean retrieval setup?",{"text":81,"@type":77},"Documents are ranked purely by the length and number of shortest substring matches to the Boolean query expressions. The method avoids supervised learning, global statistics, and term weights.",{"name":83,"@type":74,"acceptedAnswer":84},"What performance result does the agent achieve on the TREC 2024 RAG track subset?",{"text":85,"@type":77},"On 86 topics with a budget of 100 model calls per topic, the agent achieves NDCG@10 of 0.6863, placing it above many dense, sparse, and learned-sparse first-stage retrievers.","https://schema.org",{"og:url":52,"og:type":88,"og:title":13,"og:site_name":59,"og:description":14},"article",{"robots":90,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":93},[94,98,102,106,111,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":95,"show_sort_weight":96,"slug":97},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":99,"show_sort_weight":100,"slug":101},"Literature",80,"literature",{"id":53,"doc_module":4,"doc_module_name":46,"category_name":103,"show_sort_weight":104,"slug":105},"Exam",70,"exam",{"id":107,"doc_module":4,"doc_module_name":46,"category_name":108,"show_sort_weight":109,"slug":110},5,"Comic",60,"comic",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":107,"slug":138},19,"General","general"]