[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-seo-273426-105":3,"detail-sidebar-cat-0-en-105":81,"doc-detail-273426-en":131},{"code":4,"msg":5,"data":6},0,"ok",{"site_id":7,"language":8,"slug":9,"title":10,"keywords":11,"description":12,"schema_data":13,"social_meta":74,"head_meta":76,"extra_data":78,"updated_unix":80},105,"en","hallucination-span-detection-with-input-side-evidence-alignment-paper-summary","Hallucination Span Detection with Input-Side Evidence Alignment - Paper Summary","","Large language models often generate hallucinations in conditional text generation, producing unsupported or contradicted content while existing work typically judges factuality at the sentence or document level. This limits localization and prevents users from tracing which input evidence supports or contradicts specific output spans. The paper introduces hallucination span detection with input-side evidence alignment, jointly identifying hallucinated spans and aligning output tokens to corresponding input evidence. An encoder-based masked token prediction framework uses prediction confidence to distinguish faithful from hallucinated tokens, and experiments with human evaluation confirm effective detection and alignment quality.",{"@graph":14,"@context":73},[15,34,56],{"@type":16,"itemListElement":17},"BreadcrumbList",[18,23,27,31],{"item":19,"name":20,"@type":21,"position":22},"https://docshare.wps.com","Home","ListItem",1,{"item":24,"name":25,"@type":21,"position":26},"https://docshare.wps.com/document/","Document",2,{"item":28,"name":29,"@type":21,"position":30},"https://docshare.wps.com/document/research-report/","Research & Report",3,{"item":32,"name":10,"@type":21,"position":33},"https://docshare.wps.com/document/hallucination-span-detection-with-input-side-evidence-alignment-paper-summary/273426/",4,{"url":32,"name":10,"@type":35,"image":36,"author":41,"headline":10,"publisher":44,"fileFormat":47,"inLanguage":8,"description":12,"dateModified":48,"datePublished":49,"encodingFormat":47,"isAccessibleForFree":50,"interactionStatistic":51},"DigitalDocument",{"url":37,"@type":38,"width":39,"height":40},"https://docshare.wps.com/thumbnails/hallucination-span-detection-with-input-side-evidence-alignment-paper-summary/273426.png","ImageObject",300,407,{"name":42,"@type":43},"Finn","Person",{"url":19,"name":45,"@type":46},"DocShare","Organization","application/pdf","2026-10-29","2026-09-15",true,{"@type":52,"interactionType":53,"userInteractionCount":55},"InteractionCounter",{"@type":54},"ViewAction",13,{"@type":57,"mainEntity":58},"FAQPage",[59,65,69],{"name":60,"@type":61,"acceptedAnswer":62},"What problem does the proposed task address?","Question",{"text":63,"@type":64},"It detects hallucinated spans in generated text and aligns output tokens with the input evidence that supports or contradicts them, instead of only judging factuality at the sentence or document level.","Answer",{"name":66,"@type":61,"acceptedAnswer":67},"How does the method detect hallucinated tokens?",{"text":68,"@type":64},"It trains an encoder-based model to predict masked output tokens from the input representation and uses prediction confidence: faithful tokens are predictable from the input, while hallucinated tokens are not.",{"name":70,"@type":61,"acceptedAnswer":71},"Why does the approach avoid expensive manual alignment annotations?",{"text":72,"@type":64},"It is formulated under distant supervision using the intuition that faithful output tokens are predictable from their supporting context, allowing training without manually aligned input-output pairs.","https://schema.org",{"og:url":32,"og:type":75,"og:title":10,"og:site_name":45,"og:description":12},"article",{"robots":77,"canonical":32},"index,follow",{"doc_id":79,"site_id":7},273426,1789471383,{"code":4,"msg":82,"data":83},"success",[84,88,92,96,101,106,111,115,120,123,127],{"id":22,"doc_module":4,"doc_module_name":25,"category_name":85,"show_sort_weight":86,"slug":87},"Story & Novel",90,"story-novel",{"id":26,"doc_module":4,"doc_module_name":25,"category_name":89,"show_sort_weight":90,"slug":91},"Literature",80,"literature",{"id":33,"doc_module":4,"doc_module_name":25,"category_name":93,"show_sort_weight":94,"slug":95},"Exam",70,"exam",{"id":97,"doc_module":4,"doc_module_name":25,"category_name":98,"show_sort_weight":99,"slug":100},5,"Comic",60,"comic",{"id":102,"doc_module":4,"doc_module_name":25,"category_name":103,"show_sort_weight":104,"slug":105},6,"Technology",50,"technology",{"id":107,"doc_module":4,"doc_module_name":25,"category_name":108,"show_sort_weight":109,"slug":110},7,"Healthcare",40,"healthcare",{"id":112,"doc_module":4,"doc_module_name":25,"category_name":29,"show_sort_weight":113,"slug":114},8,30,"research-report",{"id":116,"doc_module":4,"doc_module_name":25,"category_name":117,"show_sort_weight":118,"slug":119},9,"Religion & Spirituality",20,"religion-spirituality",{"id":118,"doc_module":4,"doc_module_name":25,"category_name":121,"show_sort_weight":118,"slug":122},"World Cup","world-cup",{"id":124,"doc_module":4,"doc_module_name":25,"category_name":125,"show_sort_weight":124,"slug":126},10,"Lifestyle","lifestyle",{"id":128,"doc_module":4,"doc_module_name":25,"category_name":129,"show_sort_weight":97,"slug":130},19,"General","general",{"code":4,"msg":82,"data":132},{"doc_id":79,"user_id":133,"nickname":42,"user_avatar":134,"doc_module":4,"category_id":112,"category_name":29,"doc_title":10,"doc_description":12,"doc_content":135,"file_id":136,"file_url":137,"file_type":138,"file_size":139,"view_count":55,"is_deleted":4,"is_public":22,"is_downloadable":22,"audit_status":22,"page_count":124,"language":140,"language_code":8,"site_id":7,"html_lang":8,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":12,"update_tm":80,"read_time":144,"preview_page_count":4},34359740700684,"https://ap-avatar.wpscdn.com/avatar/1f400023980c374ae676?_k=1777273430885731487","Hallucination Span Detection with Input-Side Evidence Alignment  \nMiyu Yamada and Yuki Arase  \nInstitute of Science Tokyo  \n[yamada.m.ee1b@m.isct.ac.jp](yamada.m.ee1b@m.isct.ac.jp), [arase@c.titech.ac.jp](arase@c.titech.ac.jp)  \narXiv :2608 . 15804v1 [ cs .CL] 16 Aug 2026  \nAbstract  \nHallucinations remain a major obstacle to the reliable use of large language models (LLMs) in conditional text generation. Existing methods primarily assess the factuality of an entire generated text, providing limited insight into which output spans are hallucinated or how they relate to the input. We introduce the task of hallucination span detection with input-side evidence alignment, which jointly identifies hallucinated spans and aligns output tokens with the corresponding input evidence. Our approach is based on the observation that faithful output tokens are predictable from the input, whereas hallucinated tokens are not. We therefore train an encoder-based model to predict masked output tokens from the input representation, using prediction confidence for hallucination detection while naturally producing alignments to the input. Experiments show that the proposed method effectively detects hallucinated spansand identifies meaningful input-side evidence. Human evaluation confirms the quality of the predicted alignments.  \n1 Introduction  \nLarge language models (LLMs) have substantially advanced conditional text generation tasks, including text summarization and question answering. Despite these remarkable improvements, they remain prone to hallucinations, generating content that is unsupported or contradicted by the input text (Kaddour et al., 2023 ; Niu et al., 2024) . Such hallucinations undermine the reliability of LLMgenerated outputs and limit their deployment in real-world applications. To facilitate the verification of generated content, it is important not only to detect hallucinated spans in the output but also to identify the relevant portions of the input that support or contradict them. Such input–output alignment is particularly valuable for long documents, where manually tracing generated content back to its source is difficult.  \nFigure 1: Example of hallucination span detection withinput-side evidence alignment.  \nMost existing studies have formalized hallucination detection as a binary classification problem atthe sentence or document level (Jiang et al., 2024 ; Hu et al., 2024) . Although effective for estimating the factuality of an entire output, these approaches neither localize hallucinated spans nor identify the input evidence associated with individual output spans. This limits their usefulness for interpreting model predictions and assisting users in verifying generated content. To address this limitation, we introduce the task of hallucination span detection with input-side evidence alignment, which jointly identifies hallucinated spans in generated text and aligns output spans with their corresponding evidence in the input. Figure 1 illustrates the task using text summarization as an example. In the output summary, the phrase of “ADA [...] to employers with twenty or more employees” is hallucinated, because the source text states “four to fifteen”employees. An ideal model should therefore detect the hallucinated word “twenty” while identifying the relevant input span of “four to fifteen”.  \nObtaining alignments between generated text  \nand its input requires expensive manual annotation. Moreover, because LLMs evolve rapidly, constructing such annotations for every new model or domain is impractical. This motivates an approach that does not rely on manually aligned training data. We formulate the task under distant supervision, avoiding the need for manually aligned inputoutput pairs. Specifically, we exploit the observation that faithful output tokens are generally predictable from their supporting context in the input text, whereas hallucinated tokens are not. Based on this intuition, we cast hallucination span ","cbCais1eSqaAQf0j","https://ap.wps.com/l/cbCais1eSqaAQf0j","pdf",592067,"English","# Introduction\n## Hallucination Detection\n## Hallucination Span Detection with Input-Side Evidence Alignment\n# Related Work\n## Hallucination Detection","[{\"question\":\"What problem does the proposed task address?\",\"answer\":\"It detects hallucinated spans in generated text and aligns output tokens with the input evidence that supports or contradicts them, instead of only judging factuality at the sentence or document level.\"},{\"question\":\"How does the method detect hallucinated tokens?\",\"answer\":\"It trains an encoder-based model to predict masked output tokens from the input representation and uses prediction confidence: faithful tokens are predictable from the input, while hallucinated tokens are not.\"},{\"question\":\"Why does the approach avoid expensive manual alignment annotations?\",\"answer\":\"It is formulated under distant supervision using the intuition that faithful output tokens are predictable from their supporting context, allowing training without manually aligned input-output pairs.\"}]","Hallucination Span Detection with Input-Side Evidence Alignment - Paper Summary | PDF",25]