[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85300-en":3,"doc-seo-85300-105":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},85300,687197207919,"Theodora","https://ap-avatar.wpscdn.com/avatar/a000253d6f5f7c60be?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779446848396160552",6,"Technology","ProgramTab Boosting Table Reasoning of LLMs via Programmatic Paradigm","Table-based reasoning with large language models (LLMs) focuses on solving natural-language questions using structured tabular data, yet practical deployment faces key bottlenecks. Prior work degrades when tables are large due to long-text modeling difficulty and LLM input-length limits. Existing text-toSQL methods reduce table size but struggle with web tables that are often unstructured or inconsistent for SQL-style mathematical logic. ProgramTab addresses this by guiding in-context preprocessing with Python, performing focused row/column extraction, generating SQL, executing it, and producing answers. Experiments on table reasoning datasets show improved performance over LLM baselines.","ProgramTab: Boosting Table Reasoning of LLMs via Programmatic Paradigm  \nPei Guo, Enjie Liu, Yunzhi Tan* , Mochi Gao, Jianxin Zhang, Ruichao Zhong, Juntao Li, Bo Hu* , Zang Li  \n♠Big Data and AI Platform Department, Tencent, China ♢Institute of Computer Science and Technology, Soochow University, China  \n{anthonyguo, karolinaliu, boristan, mochigao, harryyfhu, [gavinzli}@tencent.com](gavinzli}@tencent.com) ;  \n[20204027008@stu.suda.edu.cn](20204027008@stu.suda.edu.cn) ; [rzhongab@connect.ust.hk](rzhongab@connect.ust.hk) ; [ljt@suda.edu.cn](ljt@suda.edu.cn)  \narXiv :2607 . 1 1207v 1 [ cs .CL] 13 Jul 2026  \nAbstract  \nTable-based reasoning with large language models (LLMs), which requires reasoning based on natural language questions and structured tabular data, has gained widespread attention. However, a series of issues still constrainthe application of this task. The previous approaches suffered from significant performance degradation when faced with large tables due to the difficulty of long text modeling and the limitation of input length for LLMs. The text-toSQL approach is used to efficiently extract key information from tables and generate smaller sub-tables. However, tabular data, especially web tables, often lack the necessary structure and consistency, making them unsuitable for performing mathematical logic operations using SQL queries. We propose the ProgramTab framework, which guides LLMs employing incontext learning to perform tabular data preprocessing with Python code, as well as the momentous contents extraction with row and column extraction and SQL generation. The experiment results on table reasoning datasets demonstrate that the ProgramTab framework effectively deals with table-based reasoning tasks and outperforms all LLM-based baselines.  \n1 Introduction  \nTables, as a popular form of data representation, play a significant role in everyday work and life. Analysis and reasoning based on tabular data have emerged as a hot topic in natural language processing, attracting wide attention from academia and industry. The main downstream tasks of tabular reasoning include table-based fact verification (Chenet al., 2020 ; Aly et al., 2021) and table-based question answering (Panupong and Percy, 2015 ; Cho et al., 2019) . The challenges of these tasks lie in how to enable language models to comprehend table data content, including text, numbers, etc.,  \n* Corresponding Author  \nTitle: 1981 Houston Oilers season  \n\n| date | opponent | result |\n| --- | --- | --- |\n| september 6, 1981 | at los angeles\u003Cbr>rams | w 27-20 |\n| september 13, 1981 | at cleveland browns | w 9-3 |\n| september 27, 1981 | miami dolphins | l 10-16 |\n| … | … | … |\n| december 20, 1981 | pittsburgh steelers | w 21-20 |\n\nFigure 1: An example of a table in WikiTQ dataset.  \nestablish their connection with user queries, and execute efficient logical reasoning and computations. Recently, LLMs (Brown et al., 2020 ; Hoffmann et al., 2022 ; OpenAI, 2022) have significantly transformed the landscape of natural language processing tasks with their impressive understanding and generation capabilities. Instead of fine-tuning the pre-trained models, sufficiently utilizing the incontext learning of LLMs to solve complex tabular data reasoning has been a mainstream direction (Chen, 2023 ; Cheng et al., 2023 ; Ye et al., 2023 ; Wang et al., 2024) . However, current methods still face several limitations. Firstly, most of the work (Cheng et al., 2023 ; Ye et al., 2023 ; Wanget al., 2024) treats the entire table as an input, which is unsuitable for tables containing large amounts of data. When the number of tokens in a table exceeds the maximum input limitation of LLMs, the content of the table will be truncated, leading to information loss and consequently affecting the performance of LLMs. This has been verified in the work of (Chen, 2023) . To mitigate the length constraint of inputs, the common approach is to utilize a programmatic language, such as generatin","cbCaioxPJQUddfl5","https://ap.wps.com/l/cbCaioxPJQUddfl5","pdf",629407,1,12,"English","en",105,"# Abstract\n# Introduction\n## Challenges in Existing Table Reasoning Methods\n## ProgramTab Framework Overview","[{\"question\":\"What problem does ProgramTab target in LLM-based table reasoning?\",\"answer\":\"It targets performance degradation caused by large tables exceeding LLM input limits and difficulties modeling long text, while also addressing unstructured or inconsistent web table formats that hinder SQL-style logic operations.\"},{\"question\":\"How does ProgramTab reduce the amount of table content fed into the LLM?\",\"answer\":\"It uses an embedding model to score relevance of table lines to the question, sorts by descending scores, and extracts the top-K relevant lines as instances instead of the full table.\"},{\"question\":\"What roles do Python and SQL play in the ProgramTab workflow?\",\"answer\":\"Python preprocessing unifies table formats and defines data types, while SQL generation and execution retrieves the most valuable information needed for answering the question.\"}]",1784202331,30,{"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},"programtab-boosting-table-reasoning-of-llms-via-programmatic-paradigm","",{"@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/technology/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/programtab-boosting-table-reasoning-of-llms-via-programmatic-paradigm/85300/",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","2026-07-17","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 ProgramTab target in LLM-based table reasoning?","Question",{"text":75,"@type":76},"It targets performance degradation caused by large tables exceeding LLM input limits and difficulties modeling long text, while also addressing unstructured or inconsistent web table formats that hinder SQL-style logic operations.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does ProgramTab reduce the amount of table content fed into the LLM?",{"text":80,"@type":76},"It uses an embedding model to score relevance of table lines to the question, sorts by descending scores, and extracts the top-K relevant lines as instances instead of the full table.",{"name":82,"@type":73,"acceptedAnswer":83},"What roles do Python and SQL play in the ProgramTab workflow?",{"text":84,"@type":76},"Python preprocessing unifies table formats and defines data types, while SQL generation and execution retrieves the most valuable information needed for answering the question.","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},[93,97,101,105,110,113,118,122,127,130,134],{"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":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":111,"slug":112},50,"technology",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":119,"doc_module":4,"doc_module_name":45,"category_name":120,"show_sort_weight":28,"slug":121},8,"Research & Report","research-report",{"id":123,"doc_module":4,"doc_module_name":45,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":45,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":45,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":45,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]