[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-657832-105":59,"doc-detail-657832-en":130},{"code":4,"msg":5,"data":6},0,"success",[7,13,18,23,28,33,38,43,48,51,55],{"id":8,"doc_module":4,"doc_module_name":9,"category_name":10,"show_sort_weight":11,"slug":12},1,"Document","Story & Novel",90,"story-novel",{"id":14,"doc_module":4,"doc_module_name":9,"category_name":15,"show_sort_weight":16,"slug":17},2,"Literature",80,"literature",{"id":19,"doc_module":4,"doc_module_name":9,"category_name":20,"show_sort_weight":21,"slug":22},4,"Exam",70,"exam",{"id":24,"doc_module":4,"doc_module_name":9,"category_name":25,"show_sort_weight":26,"slug":27},5,"Comic",60,"comic",{"id":29,"doc_module":4,"doc_module_name":9,"category_name":30,"show_sort_weight":31,"slug":32},6,"Technology",50,"technology",{"id":34,"doc_module":4,"doc_module_name":9,"category_name":35,"show_sort_weight":36,"slug":37},7,"Healthcare",40,"healthcare",{"id":39,"doc_module":4,"doc_module_name":9,"category_name":40,"show_sort_weight":41,"slug":42},8,"Research & Report",30,"research-report",{"id":44,"doc_module":4,"doc_module_name":9,"category_name":45,"show_sort_weight":46,"slug":47},9,"Religion & Spirituality",20,"religion-spirituality",{"id":46,"doc_module":4,"doc_module_name":9,"category_name":49,"show_sort_weight":46,"slug":50},"World Cup","world-cup",{"id":52,"doc_module":4,"doc_module_name":9,"category_name":53,"show_sort_weight":52,"slug":54},10,"Lifestyle","lifestyle",{"id":56,"doc_module":4,"doc_module_name":9,"category_name":57,"show_sort_weight":24,"slug":58},19,"General","general",{"code":4,"msg":60,"data":61},"ok",{"site_id":62,"language":63,"slug":64,"title":65,"keywords":66,"description":67,"schema_data":68,"social_meta":123,"head_meta":125,"extra_data":127,"updated_unix":129},105,"en","arrow-guided-vlm-enhancing-flowchart-understanding-via-arrow-direction-encoding","Arrow-Guided VLM - Enhancing Flowchart Understanding via Arrow Direction Encoding","","Flowcharts are essential for software design and business-process analysis, but vision language models often misread directional arrows and graph topology. The paper proposes a seven-stage pipeline covering arrow-aware node/endpoint detection, OCR for node text extraction, and structured prompt construction for VLM reasoning. On a 90-question benchmark from 30 annotated flowcharts, accuracy improves from 80% to 89% without task-specific VLM fine-tuning, with strongest gains on next-step queries. Limitations include detector/OCR precision dependence, a small evaluation set, and errors at multi-incoming-edge nodes.",{"@graph":69,"@context":122},[70,84,105],{"@type":71,"itemListElement":72},"BreadcrumbList",[73,77,79,82],{"item":74,"name":75,"@type":76,"position":8},"https://docshare.wps.com","Home","ListItem",{"item":78,"name":9,"@type":76,"position":14},"https://docshare.wps.com/document/",{"item":80,"name":40,"@type":76,"position":81},"https://docshare.wps.com/document/research-report/",3,{"item":83,"name":65,"@type":76,"position":19},"https://docshare.wps.com/document/arrow-guided-vlm-enhancing-flowchart-understanding-via-arrow-direction-encoding/657832/",{"url":83,"name":65,"@type":85,"image":86,"author":91,"headline":65,"publisher":94,"fileFormat":97,"inLanguage":63,"description":67,"dateModified":98,"datePublished":99,"encodingFormat":97,"isAccessibleForFree":100,"interactionStatistic":101},"DigitalDocument",{"url":87,"@type":88,"width":89,"height":90},"https://docshare.wps.com/thumbnails/arrow-guided-vlm-enhancing-flowchart-understanding-via-arrow-direction-encoding/657832.png","ImageObject",300,407,{"name":92,"@type":93},"QWQ","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-30","2026-10-26",true,{"@type":102,"interactionType":103,"userInteractionCount":19},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What problem does the paper address in flowchart understanding?","Question",{"text":112,"@type":113},"Current VLMs often misinterpret directional arrows and the graph topology that defines flowcharts, leading to incorrect reasoning.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What are the main stages of the proposed seven-stage pipeline?",{"text":117,"@type":113},"It includes arrow-aware detection of nodes and arrow endpoints, OCR to extract node text, and a coordinate-rich structured prompt that guides VLM reasoning.",{"name":119,"@type":110,"acceptedAnswer":120},"How much accuracy improvement does the method achieve and where is it strongest?",{"text":121,"@type":113},"Overall accuracy increases from 80% to 89% (+9 pp). The biggest improvement is for next-step queries, reaching 100% accuracy (30/30) from 25/30.","https://schema.org",{"og:url":83,"og:type":124,"og:title":65,"og:site_name":95,"og:description":67},"article",{"robots":126,"canonical":83},"index,follow",{"doc_id":128,"site_id":62},657832,1793146377,{"code":4,"msg":5,"data":131},{"doc_id":128,"user_id":132,"nickname":92,"user_avatar":133,"doc_module":4,"category_id":39,"category_name":40,"doc_title":65,"doc_description":67,"doc_content":134,"file_id":135,"file_url":136,"file_type":137,"file_size":138,"view_count":19,"is_deleted":4,"is_public":8,"is_downloadable":8,"audit_status":8,"page_count":139,"language":140,"language_code":63,"site_id":62,"html_lang":63,"table_of_contents":141,"faqs":142,"seo_title":143,"seo_description":67,"update_tm":144,"read_time":145},962094510011,"https://ap-avatar.wpscdn.com/davatar_155a257f0dc6eb9ab79c44ca47cae57d","arXiv :2505 .07864v 1 [ cs .AI] 9 May 2025  \nArrow-Guided VLM: Enhancing Flowchart Understanding via Arrow Direction Encoding  \nTakamitsu Omasa Ryo Koshihara Masumi Morishige  \nGalirage Inc.  \n[info@galirage.com](info@galirage.com)  \nAbstract  \nFlowcharts are indispensable tools in software design and business-process analysis, yet current Vision Language Models (VLMs) frequently misinterpret the directional arrows and graph topology that set these diagrams apart from natural images. This paper introduces a seven-stage pipeline, grouped into three broader processes—(1) arrow-aware detection of nodes and arrow endpoints; (2) Optical Character Recognition (OCR) to extract node text; and (3) construction of a structured prompt that guides the VLMs. Tested on a 90-question benchmark distilled from 30 annotated flowcharts, our method raises overall accuracy from 80% to 89%(+9 pp), a sizeable and statistically significant gain achieved without task-specific fine-tuning of the VLMs. The benefit is most pronounced for next-step queries (25/30 → 30/30; 100%, +17 pp); branch-result questions improve more modestly, and before-step queries remain difficult. A parallel evaluation with an LLM-as-aJudge protocol shows the same trends, reinforcing the advantage of explicit arrow encoding. Limitations include dependence on detector and OCR precision, the small evaluation set, and residual errors at nodes with multiple incoming edges.  \nFuture work will enlarge the benchmark with synthetic and handwritten flowchartsand assess the approach on Business Process Model and Notation (BPMN) and Unified Modeling Language (UML) .  \n1 Introduction  \nFlowcharts distill complex control flow, decision logic, and data transformations into a handful of boxes and arrows. In software engineering and business-process management, these diagrams are more than didactic artifacts, as such diagrams enable automatic code generation and serve as effective pedagogical tools [1, 2] .  \nWithin just three years, Large Language Models (LLMs) have advanced at an unprecedented pace: accuracy on the 57-subject Massive Multitask Language Understanding (MMLU) suite climbed from 43.9% with GPT-3 (2021) [3] to nearly 89% with GPT-4o [4] . VLMs likewise achieve benchmark-leading results on diverse multimodal benchmarks; for instance, GPT-4o excels on Massive Multi-discipline Multimodal Understanding (MMMU), MathVista, and Document Visual Question Answering (DocVQA) [5–8] . However, its high accuracy deteriorates markedly once explicit graph-topology reasoning is required. On the simulated subset of the FlowLearn benchmark, converting flowcharts to Mermaid code still proves challenging: on the link-level F 1 metric, Claude-3 Opus scores 0.30 and GPT-4 with Vision (GPT-4V) only 0.22 (100-sample subset; 9, Table 7), underscoring how current VLMs struggle to recover edge relationships.  \nPrevious approaches can be categorized into two main types. First, some studies couple off-theshelf object detectors such as YOLO [10] with OCR; the resulting bounding boxes and tokens are concatenated into a prompt for a VLM, yielding only modest gains over detector-free baselines.  \nPreprint.  \nFigure 1: Overview of the seven-stage pipeline: OCR, object detection, text–object fusion, arrow anchoring, node–arrow linking, graph-structured prompt generation, and VLM-based reasoning.  \nSecond, other work relies on zero-shot segmentation models, most prominently the Segment Anything Model (SAM) [11] . GenFlowchart [12], for instance, converts SAM masks into bounding boxes, adds OCR, and queries GPT-3.5 Turbo, yet still suffers from arrow-ordering ambiguities and localization noise. A complementary strand improves the detector itself—arrow-aware models like Arrow RCNN halve localization errors on handwritten diagrams [13]—but these specialised detectors have not been fused with LLMs. Their outputs feed rule-based pipelines, so branch ordering and multi-step reasoning remain unresolved. To close this gap, w","cbCaisPiGs3xpJzM","https://ap.wps.com/l/cbCaisPiGs3xpJzM","pdf",294294,11,"English","# Introduction\n## Limitations of End-to-End VLMs on Diagram Tasks\n# Related Work","[{\"question\":\"What problem does the paper address in flowchart understanding?\",\"answer\":\"Current VLMs often misinterpret directional arrows and the graph topology that defines flowcharts, leading to incorrect reasoning.\"},{\"question\":\"What are the main stages of the proposed seven-stage pipeline?\",\"answer\":\"It includes arrow-aware detection of nodes and arrow endpoints, OCR to extract node text, and a coordinate-rich structured prompt that guides VLM reasoning.\"},{\"question\":\"How much accuracy improvement does the method achieve and where is it strongest?\",\"answer\":\"Overall accuracy increases from 80% to 89% (+9 pp). The biggest improvement is for next-step queries, reaching 100% accuracy (30/30) from 25/30.\"}]","Arrow-Guided VLM - Enhancing Flowchart Understanding via Arrow Direction Encoding | PDF",1792991637,28]