[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-1-en-105":3,"doc-seo-244031-105":53,"doc-detail-244031-en":127},{"code":4,"msg":5,"data":6},0,"success",[7,14,19,24,29,34,39,44,49],{"id":8,"doc_module":9,"doc_module_name":10,"category_name":11,"show_sort_weight":12,"slug":13},11,1,"Template","Presentations",90,"presentations",{"id":15,"doc_module":9,"doc_module_name":10,"category_name":16,"show_sort_weight":17,"slug":18},12,"Resumes",80,"resumes",{"id":20,"doc_module":9,"doc_module_name":10,"category_name":21,"show_sort_weight":22,"slug":23},14,"Invoices",70,"invoices",{"id":25,"doc_module":9,"doc_module_name":10,"category_name":26,"show_sort_weight":27,"slug":28},15,"Posters",60,"posters",{"id":30,"doc_module":9,"doc_module_name":10,"category_name":31,"show_sort_weight":32,"slug":33},16,"Social Media",50,"social-media",{"id":35,"doc_module":9,"doc_module_name":10,"category_name":36,"show_sort_weight":37,"slug":38},17,"Forms",40,"forms",{"id":40,"doc_module":9,"doc_module_name":10,"category_name":41,"show_sort_weight":42,"slug":43},18,"Letters",30,"letters",{"id":45,"doc_module":9,"doc_module_name":10,"category_name":46,"show_sort_weight":47,"slug":48},21,"Paper Templates",5,"papers-templates",{"id":50,"doc_module":9,"doc_module_name":10,"category_name":51,"show_sort_weight":4,"slug":52},158,"General","general-158",{"code":4,"msg":54,"data":55},"ok",{"site_id":56,"language":57,"slug":58,"title":59,"keywords":60,"description":61,"schema_data":62,"social_meta":120,"head_meta":122,"extra_data":124,"updated_unix":126},105,"en","encoding-spreadsheets-for-large-language-models","Encoding Spreadsheets for Large Language Models","","Spreadsheets feature extensive two-dimensional grids, flexible layouts, and diverse formatting, which create major input and parsing challenges for large language models (LLMs). SHEETENCODER introduces an efficient spreadsheet encoding framework that enhances LLM understanding and reasoning. It starts from a vanilla serialization with cell addresses and formats, then addresses token constraints via structural-anchor compression, inverted-index translation, and data-format-aware aggregation. The method improves spreadsheet table detection by 25.6% and enables a fine-tuned LLM to reach 78.9% F1 while achieving an average 25× compression ratio.",{"@graph":63,"@context":119},[64,80,102],{"@type":65,"itemListElement":66},"BreadcrumbList",[67,71,74,77],{"item":68,"name":69,"@type":70,"position":9},"https://docshare.wps.com","Home","ListItem",{"item":72,"name":10,"@type":70,"position":73},"https://docshare.wps.com/template/",2,{"item":75,"name":11,"@type":70,"position":76},"https://docshare.wps.com/template/presentations/",3,{"item":78,"name":59,"@type":70,"position":79},"https://docshare.wps.com/template/encoding-spreadsheets-for-large-language-models/244031/",4,{"url":78,"name":59,"@type":81,"image":82,"author":87,"headline":59,"publisher":90,"fileFormat":93,"inLanguage":57,"description":61,"dateModified":94,"datePublished":95,"encodingFormat":93,"isAccessibleForFree":96,"interactionStatistic":97},"DigitalDocument",{"url":83,"@type":84,"width":85,"height":86},"https://docshare.wps.com/thumbnails/encoding-spreadsheets-for-large-language-models/244031.png","ImageObject",442,249,{"name":88,"@type":89},"Aurelia","Person",{"url":68,"name":91,"@type":92},"DocShare","Organization","application/pdf","2026-10-21","2026-09-12",true,{"@type":98,"interactionType":99,"userInteractionCount":101},"InteractionCounter",{"@type":100},"ViewAction",13,{"@type":103,"mainEntity":104},"FAQPage",[105,111,115],{"name":106,"@type":107,"acceptedAnswer":108},"What is SHEETENCODER and what problem does it solve?","Question",{"text":109,"@type":110},"SHEETENCODER is an encoding framework for large language models to improve spreadsheet understanding and reasoning. It addresses token limits and spreadsheet-specific representation challenges such as two-dimensional structure, cell addresses, and formatting.","Answer",{"name":112,"@type":107,"acceptedAnswer":113},"How does SHEETENCODER compress spreadsheets effectively?",{"text":114,"@type":110},"It combines three modules: structural-anchor-based compression, inverted-index translation, and data-format-aware aggregation for numerical cells. Together they reduce redundancy and token consumption while keeping the encoding lossless.",{"name":116,"@type":107,"acceptedAnswer":117},"What performance gains does SHEETENCODER achieve?",{"text":118,"@type":110},"SHEETENCODER reduces spreadsheet encoding token usage by 96%. It improves spreadsheet table detection by 25.6% over the vanilla approach in GPT-4 in-context learning, and fine-tuned models reach 78.9% F1, surpassing prior best models by 12.3%.","https://schema.org",{"og:url":78,"og:type":121,"og:title":59,"og:site_name":91,"og:description":61},"article",{"robots":123,"canonical":78},"index,follow",{"doc_id":125,"site_id":56},244031,1791093949,{"code":4,"msg":5,"data":128},{"doc_id":125,"user_id":129,"nickname":88,"user_avatar":130,"doc_module":9,"category_id":8,"category_name":11,"doc_title":59,"doc_description":61,"doc_content":131,"file_id":132,"file_url":133,"file_type":134,"file_size":135,"view_count":101,"is_deleted":4,"is_public":9,"is_downloadable":9,"audit_status":9,"page_count":45,"language":136,"language_code":57,"site_id":56,"html_lang":57,"table_of_contents":137,"faqs":138,"seo_title":139,"seo_description":61,"update_tm":140,"read_time":141},1099514068365,"https://ap-avatar.wpscdn.com/avatar/10000253d8d9f28188e?_k=1776742907772140068","Encoding Spreadsheets for Large Language Models  \nHaoyu Dong*†, Jianbo Zhao†‡, Yuzhang Tian†‡, Junyu Xiong†‡, Mengyu Zhou, Yun Lin‡, José Cambronero, Yeye He, Shi Han, Dongmei Zhang  \nMicrosoft Corporation  \nAbstract  \nSpreadsheets are characterized by their extensive two-dimensional grids, flexible layouts, and varied formatting options, which pose significant challenges for large language models (LLMs) . In response, we introduce SHEETENCODER, pioneering an efficient encoding method designed to unleash and optimize LLMs’ powerful understanding and reasoning capability on spreadsheets. Initially, we propose a vanilla serialization approach that incorporates cell addresses, values, and formats. However, this approach was limited by LLMs’ token constraints, making it impractical for most applications. To tackle this challenge, three innovative modules are proposed to compress spreadsheets effectively:  \nstructural-anchor-based compression, inverse index translation, and data-format-aware aggregation. It significantly improves performance in spreadsheet table detection task, outperforming the vanilla approach by 25.6% in GPT4’s incontext learning setting. Moreover, fine-tuned LLM with SHEETENCODER has an average compression ratio of 25×, but achieves a stateof-the-art 78.9% F1 score, surpassing the best existing models by 12.3%, demonstrating that SHEETENCODER greatly boosts LLMs’s performance on spreadsheet data.  \n1 Introduction  \nSpreadsheets are ubiquitous for data management and extensively utilized within platforms like Microsoft Excel and Google Sheets. Understanding spreadsheet layout and structure (Dong et al., 2019b ; Gol et al., 2019 ; Hulsebos et al., 2019 ; Dou et al., 2018 ; Wang et al., 2021 ; Deng et al., 2022 ; Chen and Cafarella, 2014), a longstanding challenge for traditional models and is crucial for effective data analysis, data manipulation, and various real-world applications involving intelligent  \n* Corresponding author.  \n† Equal contribution.  \n‡ Work during internship at Microsoft.  \nFigure 1: LLMs for Spreadsheet with SHEETENCODER.  \nuser interaction with spreadsheet data. Recently, the rapid development of Large Language Models (LLMs) has opened new frontiers in table processing (Li et al., 2023b) and reasoning (Cheng et al.) . However, spreadsheets pose unique challenges for LLMs due to their expansive grids that usually exceed the token limitations of popular LLMs, as well as their inherent two-dimensional layouts and structures, which are poorly suited to linear and sequential input. Furthermore, LLMs often struggle with spreadsheet-specific features such as cell addresses and formats, complicating their ability to effectively parse and utilize spreadsheet data, as detailed in Appendix A.  \nIn this paper, we introduce SHEETENCODER for LLMs, a pioneering encoding framework to unleash and maximize the potential of LLMs for spreadsheet understanding and reasoning. We initially propose a vanilla encoding method to serialize spreadsheets into sequences, augmenting the Markdown encoding method by including essential cell addresses and (optional) formats. Furthermore, large spreadsheets that exceed the token limits of LLMs not only limit their processing but also, as observed in prior studies, degrade accuracy performance as the size increases (Liu et al., 2024) . To address this challenge, we equip SHEETENCODER with three portable modules to compress spreadsheet, including:  \n20728  \nProceedings of the 2024 Conference on Empirical Methods in Natural Language Processing, pages 20728–20748 November 12-16, 2024 ©2024 Association for Computational Linguistics  \n1) Structural Anchors for Efficient Layout Understanding: Observations indicate that large spreadsheets often contain numerous homogeneous rows or columns, which contribute minimally to understanding the layout and structure (see left panel in Figure 2 (a)) . To address this, we identify structural anchors—heterogeneous rows and columns at possible ","cbCaicSrOTPlUXDc","https://ap.wps.com/l/cbCaicSrOTPlUXDc","pdf",10052476,"English","# Introduction\n## Structural Anchors for Efficient Layout Understanding\n## Inverted-Index Translation for Token Efficiency\n## Data Format Aggregation for Numerical Cells","[{\"question\":\"What is SHEETENCODER and what problem does it solve?\",\"answer\":\"SHEETENCODER is an encoding framework for large language models to improve spreadsheet understanding and reasoning. It addresses token limits and spreadsheet-specific representation challenges such as two-dimensional structure, cell addresses, and formatting.\"},{\"question\":\"How does SHEETENCODER compress spreadsheets effectively?\",\"answer\":\"It combines three modules: structural-anchor-based compression, inverted-index translation, and data-format-aware aggregation for numerical cells. Together they reduce redundancy and token consumption while keeping the encoding lossless.\"},{\"question\":\"What performance gains does SHEETENCODER achieve?\",\"answer\":\"SHEETENCODER reduces spreadsheet encoding token usage by 96%. It improves spreadsheet table detection by 25.6% over the vanilla approach in GPT-4 in-context learning, and fine-tuned models reach 78.9% F1, surpassing prior best models by 12.3%.\"}]","Encoding Spreadsheets for Large Language Models | PDF",1789211100,7]