[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86229-en":3,"doc-seo-86229-105":29,"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":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},86229,1374391974585,"Genevieve","https://ap-avatar.wpscdn.com/davatar_276721f389ce27ea32af1340a28f341c",8,"Research & Report","Cross-Architecture LLM Ensembles Feature-Based Reranking and Retrieval-Augmented Prompting for Legal Information Processing","Legal information processing covers heterogeneous retrieval, entailment, and judgment prediction problems that demand combinations of text matching, reasoning, and robust generalisation under limited supervision. The study unifies these challenges using open-weight systems across legal case retrieval, case entailment, statute retrieval, statute entailment, and legal judgment prediction. Team DU reports COLIEE 2026 results on all five tasks, with cross-architecture ensembling, feature-based reranking, and retrieval-augmented prompting delivering strong performance across differing task settings.","Cross-Architecture LLM Ensembles, Feature-Based Reranking and Retrieval-Augmented Prompting for Legal Information  \nProcessing  \nAmal Saad Alshehri  \nDurham University Department of Computer Science Durham, United Kingdom Jazan University  \nDepartment of Computer Science Jazan, Saudi Arabia [ashahri@jazanu.edu.sa](ashahri@jazanu.edu.sa)  \nNelly Bencomo  \nDurham University Department of Computer Science Durham, United Kingdom [nelly.bencomo@durham.ac.uk](nelly.bencomo@durham.ac.uk)  \nAmir Atapour-Abarghouei  \nDurham University Department of Computer Science Durham, United Kingdom amir.atapour[abarghouei@durham.ac.uk](abarghouei@durham.ac.uk)  \narXiv :2607 . 1 1400v 1 [ cs .CL] 13 Jul 2026  \nABSTRACT  \nLegal information processing encompasses a heterogeneous set of retrieval, entailment and judgment prediction problems, requiring combinations of text matching, reasoning and robust generalisation with limited supervision. This paper presents a unified study of these challenges through a suite of open-weight systems spanning legal case retrieval, case entailment, statute retrieval and entailment and legal judgment prediction. In this paper, we report Team DU’s participation in all five tasks of COLIEE 2026, with all systems relying exclusively on open-weight models. For Tasks 3 and 4, all models were released before 15 July 2025, as required by the competition rules. For Task 4 (statute entailment), a cross-architecture ensemble of nine models from three families achieves 96 .3% accuracy, placing first among 33 submissions from 11 teams. For the Pilot Task (tort prediction and rationale extraction), a multi-view system that combines five claim-level models and refines the case verdict using features derived from the claim predictions achieves 73.1% TP accuracy and 68.2% RE F1 as an unofficial submission, scoring above all official entries on TP and matching the highest on RE. For Task 2 (legal case entailment), changing only the prompt instruction from single-selection to multi-selection raises F1 from 0.343 to 0.555 in post-competition evaluation on released gold labels, exceeding the best official submission (F1 = 0.490). For Task 3 (statute retrieval andentailment), replacing the entailment model with Qwen3-235B anda structured legal reasoning prompt raises accuracy from 79.3% to 91.5% in post-competition analysis. For Task 1 (legal case retrieval), a learning-to-rank system that combines lexical and semantic retrieval with structural, citation authority, and temporal features (34 in total) achieves F1 = 0.314 (rank 11 of 54 submissions from 22 teams) . Taken together, these results show that legal information processing benefits from different forms of inductive bias across tasks, with cross-architecture ensembling, feature-based reranking and retrieval-augmented prompting each proving most effective indifferent settings.  \nPermission to make digital or hard copies of part or all ofthis work for personal or classroom use is granted without fee provided that copies are not made or distributed for profit or commercial advantage and that copies bear this notice and the full citation on the first page. Copyrights for third-party components of this work must be honored. For all other uses, contact the owner/author(s) .  \nCOLIEE 2026, June 12, 2026, Singapore  \n© 2026 Copyright held by the owner/author(s) .  \nKEYWORDS  \nCOLIEE, legal information retrieval, legal entailment, LLM ensemble, learning-to-rank, legal judgment prediction  \n1 INTRODUCTION  \nLegal information processing involves a diverse family of problems, including document retrieval, textual entailment and judgment prediction, each of which places different demands on representation learning, structured reasoning and decision-making under domainspecific constraints [1]. These tasks become particularly challenging in legal settings, where documents are long, terminology is highly specialised, relevant evidence may be distributed unevenly across a text and small linguistic","cbCaiei95O2BRFij","https://ap.wps.com/l/cbCaiei95O2BRFij","pdf",600632,1,10,"English","en",105,"# Introduction\n# Tasks and Benchmark (COLIEE 2026)\n## Team DU Participation and Results\n# System Approaches (by Task)\n## Task 1: Legal Case Retrieval\n## Task 2: Legal Case Entailment\n## Task 3: Statute Retrieval and Entailment\n## Task 4: Statute Entailment\n## Pilot Task: Tort Prediction and Rationale Extraction","[{\"question\":\"What kinds of problems does the paper address in legal information processing?\",\"answer\":\"It addresses legal information retrieval, textual entailment (case and statute), and legal judgment prediction, each requiring different representation and reasoning demands.\"},{\"question\":\"Which benchmark and competition does the study use?\",\"answer\":\"The paper evaluates methods through COLIEE 2026, which includes four established tasks plus one pilot task.\"},{\"question\":\"What approach led to the strongest performance for Task 4 and the Pilot Task?\",\"answer\":\"Task 4 uses a cross-architecture ensemble of nine open-weight models from three families, while the Pilot Task uses a multi-view system combining claim-level models and refining the verdict using features from claim predictions.\"}]",1784209649,25,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":27},"cross-architecture-llm-ensembles-feature-based-reranking-and-retrieval-augmented-prompting-for-legal-information-processing","",{"@graph":35,"@context":85},[36,53,68],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/cross-architecture-llm-ensembles-feature-based-reranking-and-retrieval-augmented-prompting-for-legal-information-processing/86229/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-27","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 kinds of problems does the paper address in legal information processing?","Question",{"text":75,"@type":76},"It addresses legal information retrieval, textual entailment (case and statute), and legal judgment prediction, each requiring different representation and reasoning demands.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"Which benchmark and competition does the study use?",{"text":80,"@type":76},"The paper evaluates methods through COLIEE 2026, which includes four established tasks plus one pilot task.",{"name":82,"@type":73,"acceptedAnswer":83},"What approach led to the strongest performance for Task 4 and the Pilot Task?",{"text":84,"@type":76},"Task 4 uses a cross-architecture ensemble of nine open-weight models from three families, while the Pilot Task uses a multi-view system combining claim-level models and refining the verdict using features from claim predictions.","https://schema.org",{"og:url":51,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,134],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":21,"slug":133},"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]