[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-81504-en":3,"doc-seo-81504-105":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},81504,1099513958607,"Jiven","https://ap-avatar.wpscdn.com/avatar/100002390cf8733938c?x-image-process=image/resize,m_fixed,w_180,h_180&k=1778829742770036399",8,"Research & Report","IFAR: Multi-Perspective and Multi-Level Causal Discovery with LLMs","Large language models (LLMs) have advanced rapidly, yet abductive reasoning—inferring causes for observed phenomena—remains underexplored, especially when causes are multi-perspective and multi-level. The work introduces the DeepAbduction dataset for pollution and disease cause tracing and presents the INVERSE-FORWARD ABDUCTIVE REASONING (IFAR) framework. IFAR is zero-shot, using generalized backward abduction plus relation-by-relation forward verification, improving F1 by about 40% while preserving recall-precision balance and boosting even non-reasoning LLMs. Code is planned after acceptance.","IFAR: Multi-Perspective and Multi-Level Causal  \nDiscovery with LLMs  \nJinwei He and Feng Lu  \narXiv :2409 .05559v 3 [ cs .AI] 10 Jul 2026  \nAbstract—Large language models (LLMs) have developed rapidly, and their reasoning capabilities have become a hot research topic. However, there is still limited exploration of abductive reasoning. The multi-perspective and multi-level of causes is one of the core challenges of abductive reasoning, which cannot be solved well by existing methods. We construct a specialized dataset named DeepAbduction, which is designed for tracing the causes of pollution and disease, addressing the lack of datasets in this field. We propose INVERSE-FORWARD ABDUCTIVE REASONING (IFAR) framework for LLMs multiperspective and multi-level abductive reasoning. IFAR is zeroshot and combines generalized backward reasoning with relationby-relation forward verification. Experimental results show that IFAR achieves an improvement of approximately 40% in the F1 score compared to other methods under mainstream LLMs, while maintaining a balance between recall and precision. Furthermore, IFAR enhances the performance of non-reasoning LLMs to surpass LLMs which have been trained for reasoning, and remains effective when applied to the latter. Code will be released after the acceptance of our work.  \nImpact Statement—Abductive reasoning is essential for understanding complex real-world phenomena such as pollution sources and disease transmission. However, existing language models struggle with this capability, limiting their usefulness in these analytical tasks. Our dataset and framework can improve the reliability of abductive reasoning in LLMs and significantly enhances their accuracy in tracing multi-level and multi-perspective causes. This advancement enables AI systems to better support applications in environmental monitoring, public health analysis, scientific investigation, and other domains that rely on causal understanding. By reducing expert workload and improving decision quality, our method has the potential to be applied in critical societal and industrial settings.  \nIndex Terms—Large Language Models, Multi-Perspective and Multi-Level Abductive Reasoning, Large Language Models Reasoning, Abductive Datasets  \nI. INTRODUCTION  \nTHE development of large language models (LLMs) marks  \na pivotal step toward artificial general intelligence (AGI), with research focusing on fine-tuning [1], [2], agents [3], [4],[5], and applications [6], [7], [8] . But LLMs still fall short of true intelligence: deep reasoning ability. As a result, recent training paradigms have focused on enhancing reasoning, and developing methods to further enhance LLMs reasoning capacity has become a key research focus [9] .  \nCurrent work is categorized into prompt-based methods (e.g., CoT [10], ToT [11]), which can be directly used yet  \nJinwei He and Feng Lu are with State Key Laboratory of VR Technology and Systems, School of CSE, Beihang University (e-mail: [lufeng@buaa.edu.cn](lufeng@buaa.edu.cn)).  \nCorresponding author: Feng Lu.  \nThis work is available as a preprint on arXiv.  \n\n| \u003Cbr>Multiple Perspectives (e.g., Physiology, Habits and Mentality) and Multiple Levels (Direct / Indirect / Root)： |  |  |\n| --- | --- | --- |\n| Phenomenon:\u003Cbr>Hypertension | Indirect Cause:\u003Cbr>(Habits)\u003Cbr>Lack of Sleep | Root Cause:\u003Cbr>(Mentality)\u003Cbr>Long Term Pressure |\n\nFig. 1: An example for multi-perspective and multi-level abductive reasoning. Abductive problem for previous study fails to distinguish the perspectives and reasoning levels of the causes. For eaxmple, causes of hypertension have different levels, including direct cause, indirect cause and root cause. They are also in different perspectives, such as physiology, habits and mentality perspectives.  \noffer limited performance gains, and reinforcement learning or fine-tuning, which achieve the state-of-the-art reasoning performance but at the cost of substantial data and computational resources.","cbCaibYKki6ecMCg","https://ap.wps.com/l/cbCaibYKki6ecMCg","pdf",3063351,3,1,11,"English","en",105,"# Introduction\n## Background and Motivation\n## Challenges of Multi-Perspective, Multi-Level Abduction\n## Dataset: DeepAbduction\n## Framework: IFAR","[{\"question\":\"What problem does IFAR target in LLM reasoning?\",\"answer\":\"IFAR targets abductive reasoning where causes must be identified across multiple perspectives and multiple levels, which existing methods and datasets do not address well.\"},{\"question\":\"What is DeepAbduction and what is it designed for?\",\"answer\":\"DeepAbduction is a specialized dataset built to support tracing causes of pollution and disease, focusing on multi-perspective and multi-level causal structures in long-text settings.\"},{\"question\":\"How does the IFAR framework work and what benefits does it show?\",\"answer\":\"IFAR combines generalized backward reasoning (inverse) with relation-by-relation forward verification. Experiments report an approximate 40% F1 improvement over other methods under mainstream LLMs while maintaining a balance between recall and precision.\"}]",1784173853,28,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"ifar-multi-perspective-and-multi-level-causal-discovery-with-llms","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"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":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/ifar-multi-perspective-and-multi-level-causal-discovery-with-llms/81504/",4,{"url":51,"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","2026-07-24","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 IFAR target in LLM reasoning?","Question",{"text":75,"@type":76},"IFAR targets abductive reasoning where causes must be identified across multiple perspectives and multiple levels, which existing methods and datasets do not address well.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is DeepAbduction and what is it designed for?",{"text":80,"@type":76},"DeepAbduction is a specialized dataset built to support tracing causes of pollution and disease, focusing on multi-perspective and multi-level causal structures in long-text settings.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the IFAR framework work and what benefits does it show?",{"text":84,"@type":76},"IFAR combines generalized backward reasoning (inverse) with relation-by-relation forward verification. Experiments report an approximate 40% F1 improvement over other methods under mainstream LLMs while maintaining a balance between recall and precision.","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":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"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":52,"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":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":106,"slug":138},19,"General","general"]