[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86034-en":3,"doc-seo-86034-105":30,"detail-sidebar-cat-0-en-105":91},{"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},86034,1099514067438,"River Wang","https://ap-avatar.wpscdn.com/avatar/100002539ee87300030?x-image-process=image/resize,m_fixed,w_180,h_180&k=1780474512215547542",8,"Research & Report","STEC Evidence Compression for Deep Search in Open-domain Multi-Hop QA","Open-domain multi-hop question answering with LLM-based search agents combines retrieval and reasoning but generates multiple search trajectories whose evidence is heterogeneous, redundant, incomplete, or conflicting, making reliable final answer selection difficult. Directly comparing raw trajectories exposes the verifier to noisy and unaligned content, while comparing answer strings ignores evidence quality. STEC proposes an evidence compression framework that groups trajectories by normalized answer identity, builds answer-specific evidence representations, and performs evidence-guided answer verification to select the final candidate.","arXiv :2607 . 10795v 1 [ cs .AI] 12 Jul 2026  \nSTEC: Evidence Compression for Deep Search in Open-domain Multi-Hop QA  \nXinkang Li 1 , Rong Jiang 1 (􀀌), Xin Song 1 , Ye Wang2 , Yue Han 1 , and Changjian  \nLi 1  \n1 National University of Defense Technology, China [xinkangli@nudt.edu.cn](xinkangli@nudt.edu.cn) , [jiangrong@nudt.edu.cn](jiangrong@nudt.edu.cn) , [songxin@nudt.edu.cn](songxin@nudt.edu.cn) ,  \n[hanyue@nudt.edu.cn](hanyue@nudt.edu.cn) , [lichangjian23@nudt.edu.cn](lichangjian23@nudt.edu.cn)  \n2 Harbin Institute of Technology, Shenzhen, China  \n[wangye2020@hit.edu.cn](wangye2020@hit.edu.cn)  \nAbstract. In open-domain multi-hop question answering (QA), LLMbased search agents offer a promising approach to knowledge-intensive QA by combining retrieval with reasoning. Existing methods mainly improve open-domain multi-hop QA through reasoning paradigms, retrieval interaction, and search strategy optimization. However, using multiple search trajectories introduces a challenging final answer selection problem. Different trajectories may support different candidates, and the retrieved information can be heterogeneous, redundant, incomplete, or conflicting. Directly comparing raw trajectories exposes the verifier to noisy and unaligned content, while comparing answer strings ignores the evidence supporting each candidate, making reliable final selection difficult.  \nTo address this challenge, we propose STEC, an evidence compression framework for final answer selection in multi-hop QA. STEC selects the final answer from the existing candidate set through two mechanisms: (1) Answer-Level Evidence Compression, which groups trajectories by normalized answer identity and converts each answer group into a candidatespecific evidence representation; and (2) Evidence-Guided Answer Verification, which compares these representations and selects the final answer from the candidate set. The design shifts final selection from raw trajectory comparison to candidate-level evidence comparison. We evaluate STEC on four open-domain multi-hop QA benchmarks against representative baselines. Experimental results show that STEC performs best overall among the compared methods, and ablation results provide evidence that answer-level evidence compression contributes to final answer selection.  \nKeywords: Deep Search · Multi-Hop Question Answering · Search Trajectory · Evidence Compression · Answer Verification  \n1 Introduction  \nOpen-domain multi-hop question answering is a challenging knowledge-intensive task. A system needs to retrieve relevant information from large-scale open corpora and integrate it across entities, relations, and events to support coherent  \n2 X. Li et al.  \n7[KYZOUT  \n\n| 2GXMK􀀓YIGRK 5VKT )UXV[Y |  |  |  |\n| --- | --- | --- | --- |\n|  |  |  |  |\n|  |  | Ș |  |\n\n'TY]KX  \nFig. 1. Illustration of open-domain multi-hop question answering. A system retrieves relevant information from a large-scale open corpus and reasons across multiple hops to derive the final answer.  \nmulti-step reasoning. Unlike single-hop question answering, a multi-hop question usually cannot be answered with a single retrieved passage. The model must combine information from multiple sources through reliable multi-step reasoning, and retrieval omissions, information noise, or intermediate reasoning errors can all lead to an incorrect final answer. Fig. 1 illustrates this task setting.  \nTraditional retrieval-augmented generation provides a basic paradigm for knowledge-intensive question answering. The model first retrieves external context and then generates an answer based on the retrieved results, which mitigates knowledge gaps and factual errors that arise when LLMs rely only on parametric knowledge [1] . However, open-domain multi-hop question answering often involves entities, relations, and constraints that emerge gradually during reasoning. As reasoning proceeds, the model may need to resolve new intermediate entities, compare different facts, or verify conditions introduc","cbCaiqKEmGqTLqKK","https://ap.wps.com/l/cbCaiqKEmGqTLqKK","pdf",379410,4,1,16,"English","en",105,"# Introduction\n## Open-domain multi-hop QA and retrieval limits\n## Deep search as interleaved search and reasoning\n## Final answer selection challenge","[{\"question\":\"What problem does STEC address in open-domain multi-hop QA?\",\"answer\":\"STEC targets the final answer selection difficulty caused by multiple deep-search trajectories producing conflicting or incomplete evidence for different candidates.\"},{\"question\":\"How does STEC compress evidence for final answer selection?\",\"answer\":\"STEC groups trajectories by normalized answer identity and converts each answer group into a candidate-specific evidence representation.\"},{\"question\":\"How does STEC choose the final answer among candidates?\",\"answer\":\"STEC uses evidence-guided answer verification by comparing the compressed evidence representations and selecting the final answer from the candidate set.\"}]",1784207987,40,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"stec-evidence-compression-for-deep-search-in-open-domain-multi-hop-qa","",{"@graph":36,"@context":85},[37,53,68],{"@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":20},"https://docshare.wps.com/document/stec-evidence-compression-for-deep-search-in-open-domain-multi-hop-qa/86034/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-26","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81],{"name":72,"@type":73,"acceptedAnswer":74},"What problem does STEC address in open-domain multi-hop QA?","Question",{"text":75,"@type":76},"STEC targets the final answer selection difficulty caused by multiple deep-search trajectories producing conflicting or incomplete evidence for different candidates.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does STEC compress evidence for final answer selection?",{"text":80,"@type":76},"STEC groups trajectories by normalized answer identity and converts each answer group into a candidate-specific evidence representation.",{"name":82,"@type":73,"acceptedAnswer":83},"How does STEC choose the final answer among candidates?",{"text":84,"@type":76},"STEC uses evidence-guided answer verification by comparing the compressed evidence representations and selecting the final answer from the candidate set.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":29,"slug":118},7,"Healthcare","healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]