[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-1-en-105":3,"doc-detail-564095-en":53,"doc-seo-564095-105":76},{"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":5,"data":54},{"doc_id":55,"user_id":56,"nickname":57,"user_avatar":58,"doc_module":9,"category_id":50,"category_name":51,"doc_title":59,"doc_description":60,"doc_content":61,"file_id":62,"file_url":63,"file_type":64,"file_size":65,"view_count":66,"is_deleted":4,"is_public":9,"is_downloadable":9,"audit_status":9,"page_count":67,"language":68,"language_code":69,"site_id":70,"html_lang":69,"table_of_contents":71,"faqs":72,"seo_title":73,"seo_description":60,"update_tm":74,"read_time":75},564095,3985747862462,"dbqoiu","https://ap-avatar.wpscdn.com/davatar_3d24733baf745e90a7e4bdd5f77d97b2","Follow the Flow - Fine-grained Flowchart Attribution with Neurosymbolic Agents","Flowcharts visualize decision-making, but their non-linear structure and intricate visual-text links hinder LLM interpretation, leading vision-language models to hallucinate nonexistent connections and decision paths. This undermines reliability in high-stakes automated flowchart reasoning. The paper introduces Fine-grained Flowchart Attribution, tracing specific flowchart components that ground an LLM response to improve verifiability and explainability. It presents FlowPathAgent, a neurosymbolic graph-based agent, and FlowExplainBench to evaluate attributions across styles and question types; results show 10–14% gains while reducing visual hallucinations.","Follow the Flow: Fine-grained Flowchart Attribution with Neurosymbolic Agents  \nManan Suri , Puneet Mathur  * , Nedim Lipka ,  \nFranck Dernoncourt , Ryan A. Rossi  , Vivek Gupta , Dinesh Manocha  \nUniversity of Maryland  Adobe Research  ASU  \n[manans@umd.edu](manans@umd.edu), [puneetm@adobe.com](puneetm@adobe.com), [vgupt140@asu.edu](vgupt140@asu.edu)  \nAbstract  \nFlowcharts are a critical tool for visualizing decision-making processes. However, their non-linear structure and complex visual-textual relationships make it challenging to interpret them using LLMs, as vision-language models frequently hallucinate nonexistent connectionsand decision paths when analyzing these diagrams. This leads to compromised reliability for automated flowchart processing in critical domains such as logistics, health, and engineering. We introduce the task of Fine-grained Flowchart Attribution, which traces specific components grounding a flowchart referring LLM response. Flowchart Attribution ensures the verifiability of LLM predictions and improves explainability by linking generated responses to the flowchart’s structure. We propose FlowPathAgent, a neurosymbolic agent that performs fine-grained post hoc attribution through graph-based reasoning. It first segments the flowchart, then converts it into a structured symbolic graph, and then employs an agentic approach to dynamically interact with the graph, to generate attribution paths. Additionally, we present FlowExplainBench, a novel benchmark for evaluating flowchart attributions across diverse styles, domains, and question types. Experimental results show that FlowPathAgent mitigates visual hallucinations in LLM answers over flowchart QA, outperforming strong baselines by 10-14% on our proposed FlowExplainBench dataset. 1  \n1 Introduction  \nFlowcharts are a fundamental tool for representing structured decision-making processes. Used across domains such as software engineering, business process modeling, and instructional design, flowcharts provide a visual roadmap of logical operations, guiding both human users and automated systems (Charntaweekhun and Wangsiripi-  \n*Primary Research Mentor  \n1Code and Data: MananSuri27/FollowTheFlow  \nWhat is the immediate nextstep after utilizing prepared items for seeking help, and what decision led to this step?  \nHallucinated Response  \nF lowchart  \nThe immediate next step is notifying trusted contact of travel plans, and this step was motivated by a positive response to the need to leave the vehicle  \nAttributed path is not logically consistent, and offers opportunity to eliminate this response.  \nFlowPathAgent  \nYes  \nNo  \nFigure 1: Attribution (represented by •−•−•) with FlowPathAgent ensures logical consistency in flowchart-based reasoning. FlowPathAgent uses a neurosymbolic approach to generate attribution paths ( ➊ & ➋) in the flowchart. This enhances interpretability and reliability in flowchart driven automated decision-making.  \ntak, 2006 ; Perols and Perols, 2024 ; Zimmermann et al., 2024 ; Ensmenger, 2016) . Their structured yet visual nature makes them an effective medium for conveying procedural logic. However, interpreting flowcharts accurately is challenging due to their nonlinear structures (branching and loopbased control flow), where meaning emerges from the interplay between textual content, visual arrangement, and logical dependencies. Ambiguities in flowchart interpretation arise from diverse notational conventions, implicit relationships, and misinferred steps, making precise attribution of  \n22485  \nProceedings of the 2025 Conference on Empirical Methods in Natural Language Processing, pages 22485–22508 November 4-9, 2025 ©2025 Association for Computational Linguistics  \ninformation sources difficult (Eppler et al., 2008) . Recent advancements in Vision Language Models (VLMs) have enabled substantial progress in flowchart processing (Singh et al., 2024) . These models leverage both textual and visual information, allowing them to","cbCainkSQYPz1PL3","https://ap.wps.com/l/cbCainkSQYPz1PL3","pdf",8190990,7,24,"English","en",105,"# Abstract\n# Introduction\n## Background on flowcharts and hallucination\n## Fine-grained flowchart attribution task\n## Proposed approach: FlowPathAgent\n## Benchmark: FlowExplainBench","[{\"question\":\"Why is fine-grained attribution important for flowchart understanding by LLMs?\",\"answer\":\"Flowcharts combine textual content and visual layout in nonlinear ways, and models may produce answers not grounded in the diagram. Fine-grained attribution traces the specific diagram components that support an LLM response, improving verifiability and explainability.\"},{\"question\":\"What is FlowPathAgent and what does it do?\",\"answer\":\"FlowPathAgent is a neurosymbolic agent that performs fine-grained post hoc attribution. It segments a flowchart, converts it into a structured symbolic graph, and uses graph-based reasoning to generate attribution paths.\"},{\"question\":\"How is FlowExplainBench used in the paper?\",\"answer\":\"FlowExplainBench is a benchmark for evaluating flowchart attributions across diverse styles, domains, and question types. Experimental results show FlowPathAgent mitigates visual hallucinations and outperforms strong baselines by 10–14% on the benchmark dataset.\"}]","Follow the Flow - Fine-grained Flowchart Attribution with Neurosymbolic Agents | PDF",1791840232,8,{"code":4,"msg":77,"data":78},"ok",{"site_id":70,"language":69,"slug":79,"title":59,"keywords":80,"description":60,"schema_data":81,"social_meta":137,"head_meta":139,"extra_data":141,"updated_unix":142},"follow-the-flow-fine-grained-flowchart-attribution-with-neurosymbolic-agents","",{"@graph":82,"@context":136},[83,99,119],{"@type":84,"itemListElement":85},"BreadcrumbList",[86,90,93,96],{"item":87,"name":88,"@type":89,"position":9},"https://docshare.wps.com","Home","ListItem",{"item":91,"name":10,"@type":89,"position":92},"https://docshare.wps.com/template/",2,{"item":94,"name":51,"@type":89,"position":95},"https://docshare.wps.com/template/general/",3,{"item":97,"name":59,"@type":89,"position":98},"https://docshare.wps.com/template/follow-the-flow-fine-grained-flowchart-attribution-with-neurosymbolic-agents/564095/",4,{"url":97,"name":59,"@type":100,"image":101,"author":106,"headline":59,"publisher":108,"fileFormat":111,"inLanguage":69,"description":60,"dateModified":112,"datePublished":113,"encodingFormat":111,"isAccessibleForFree":114,"interactionStatistic":115},"DigitalDocument",{"url":102,"@type":103,"width":104,"height":105},"https://docshare.wps.com/thumbnails/follow-the-flow-fine-grained-flowchart-attribution-with-neurosymbolic-agents/564095.png","ImageObject",442,249,{"name":57,"@type":107},"Person",{"url":87,"name":109,"@type":110},"DocShare","Organization","application/pdf","2026-10-29","2026-10-12",true,{"@type":116,"interactionType":117,"userInteractionCount":66},"InteractionCounter",{"@type":118},"ViewAction",{"@type":120,"mainEntity":121},"FAQPage",[122,128,132],{"name":123,"@type":124,"acceptedAnswer":125},"Why is fine-grained attribution important for flowchart understanding by LLMs?","Question",{"text":126,"@type":127},"Flowcharts combine textual content and visual layout in nonlinear ways, and models may produce answers not grounded in the diagram. Fine-grained attribution traces the specific diagram components that support an LLM response, improving verifiability and explainability.","Answer",{"name":129,"@type":124,"acceptedAnswer":130},"What is FlowPathAgent and what does it do?",{"text":131,"@type":127},"FlowPathAgent is a neurosymbolic agent that performs fine-grained post hoc attribution. It segments a flowchart, converts it into a structured symbolic graph, and uses graph-based reasoning to generate attribution paths.",{"name":133,"@type":124,"acceptedAnswer":134},"How is FlowExplainBench used in the paper?",{"text":135,"@type":127},"FlowExplainBench is a benchmark for evaluating flowchart attributions across diverse styles, domains, and question types. Experimental results show FlowPathAgent mitigates visual hallucinations and outperforms strong baselines by 10–14% on the benchmark dataset.","https://schema.org",{"og:url":97,"og:type":138,"og:title":59,"og:site_name":109,"og:description":60},"article",{"robots":140,"canonical":97},"index,follow",{"doc_id":55,"site_id":70},1792453510]