[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-82428-en":3,"doc-seo-82428-105":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},82428,7971461741311,"Ophelia","https://ap-avatar.wpscdn.com/avatar/74000253aff267980c6?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779345379180704826",8,"Research & Report","Task-Specific Multimodal Question Answering Agents via Confidence Calibration and Incremental Reasoning for QANTA 2026","Submission for the QANTA 2026 shared challenge targets efficient multimodal question answering in a quizbowl setting with incrementally revealed text and images under realistic efficiency constraints. The work distinguishes Tossup and Bonus objectives: early calibrated commitment versus accurate selection with human usefulness. A task-specific two-agent architecture uses a confidence-calibrated numeric reasoning policy for Tossups and leadin-aware structured relational multimodal evidence integration for Bonuses, avoiding retrieval pipelines and ensembles. The system attains an overall leaderboard score of 0.402.","Task-Specific Multimodal Question Answering Agents via Confidence Calibration and Incremental Reasoning for QANTA 2026  \nNirjhar Das 1 Md. Al-Mamun Provath 1  \narXiv :2607 .09623v 1 [ cs .CL] 10 Jul 2026  \nAbstract  \nWe present our submission to the QANTA 2026 shared challenge at the ICML 2026 Workshop on Efficient Multimodal Question Answering (EMMQA) . Quanta evaluates multimodal quizbowl systems that answer pyramid-style questions from incrementally revealed text and accompanying images while operating under realistic efficiency constraints. The challenge consists of two distinct tasks: Tossup questions, which require deciding when to answer under uncertainty, and Bonus questions, which emphasize accurate answer selection and human adoption. To address these differing objectives, we develop a task-specific two-agent architecture. Our Tossup agent utilizesa GPT-4o-mini-class model (referred to as GPT- 4.1-mini in the competition logs) with confidencecalibrated answering and a domain-specific numeric reasoning policy that reduces overconfident predictions from isolated quantitative clues. Our Bonus agent uses GPT-4o-class model (referred to as GPT-4.1) with leadin-aware reasoning, structured relational reasoning, and multimodal evidence integration to improve exact answer selection. Rather than relying on a retrieval pipeline or model ensembles, our approach emphasizes efficient reasoning policies and confidence calibration within a hosted-only environment. Our system achieved the highest overall leaderboard score of 0.402, including a Tossup score of 0.238 and a Bonus Effect score of 0 .164. The results demonstrate that lightweight, task-specific reasoning strategies can provide strong performance on resource-constrained multimodal question answering benchmarks.  \n1Department of Computer Science and Engineering, Chittagong University of Engineering & Technology, Chattogram, Bangladesh. Correspondence to: Nirjhar Das \u003C[nirjhardasami@gmail.com](nirjhardasami@gmail.com) >.  \nAccepted at EMM-QA 2026 Workshop, ICML 2026 (Non-Archival; Published on Workshop Website)  \n1. Introduction  \nThe QANTA 2026 shared challenge studies efficient multimodal question answering in the quizbowl setting, where systems answer pyramid-style questions from incrementally revealed text and accompanying images. Multimodal question answering has been extensively studied in visual question answering and multimodal reasoning settings (Antol et al., 2015) . However, quizbowl-style question answering presents a unique challenge because systems must continuously update their beliefs and decide when sufficient evidence exists to commit to an answer (Boyd-Graberet al., 2012 ; Wallace et al., 2019) . Unlike conventional question answering benchmarks that provide complete context, quizbowl requires systems to reason under uncertainty and determine when sufficient evidence exists to commit to an answer. This setting makes confidence calibration and efficient reasoning as important as answer accuracy.  \nQANTA consists of two complementary tasks. Tossup questions require systems to decide when to answer as clues are revealed, balancing the benefit of early correct answers against penalties for incorrect predictions. Bonus questions provide complete context and emphasize accurate answer selection, calibration, and usefulness to human teammates. The inclusion of multimodal evidence further requires systems to integrate textual and visual information while remaining computationally efficient.  \nThe distinct objectives of the two tasks motivate a taskspecific design. We develop a two-agent architecture consisting of a GPT-4.1-mini-based Tossup agent and a GPT- 4.1- based Bonus agent. The Tossup agent focuses on confidence-calibrated answering and robust quantitative reasoning, while the Bonus agent emphasizes leadin-aware reasoning, structured answer selection, and multimodal evidence integration. Our system achieved the highest overall leaderboard score of 0.402, incl","cbCaim1KpQThcane","https://ap.wps.com/l/cbCaim1KpQThcane","pdf",299548,1,10,"English","en",105,"# Abstract\n# Introduction\n# Task Overview\n## Tossup Task","[{\"question\":\"What are the two tasks in QANTA 2026 and how do their goals differ?\",\"answer\":\"QANTA 2026 includes Tossup and Bonus tasks. Tossups require deciding when to answer under uncertainty as clues are revealed, while Bonuses emphasize accurate answer selection and human usefulness with complete context.\"},{\"question\":\"How does the submitted system handle confidence calibration for Tossup questions?\",\"answer\":\"The Tossup agent uses a confidence-calibrated answering approach with a domain-specific numeric reasoning policy designed to reduce overconfident predictions from isolated quantitative clues.\"},{\"question\":\"What reasoning approach does the Bonus agent use to improve answer selection?\",\"answer\":\"The Bonus agent applies leadin-aware reasoning, structured relational reasoning, and multimodal evidence integration to select the exact answer more effectively.\"}]",1784180330,25,{"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},"task-specific-multimodal-question-answering-agents-via-confidence-calibration-and-incremental-reasoning-for-qanta-2026","",{"@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/task-specific-multimodal-question-answering-agents-via-confidence-calibration-and-incremental-reasoning-for-qanta-2026/82428/",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 are the two tasks in QANTA 2026 and how do their goals differ?","Question",{"text":75,"@type":76},"QANTA 2026 includes Tossup and Bonus tasks. Tossups require deciding when to answer under uncertainty as clues are revealed, while Bonuses emphasize accurate answer selection and human usefulness with complete context.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does the submitted system handle confidence calibration for Tossup questions?",{"text":80,"@type":76},"The Tossup agent uses a confidence-calibrated answering approach with a domain-specific numeric reasoning policy designed to reduce overconfident predictions from isolated quantitative clues.",{"name":82,"@type":73,"acceptedAnswer":83},"What reasoning approach does the Bonus agent use to improve answer selection?",{"text":84,"@type":76},"The Bonus agent applies leadin-aware reasoning, structured relational reasoning, and multimodal evidence integration to select the exact answer more effectively.","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"]