[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-273436-105":59,"doc-detail-273436-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","seer-long-context-reasoning-via-selective-visual-text-compression","SEER - Long-Context Reasoning via Selective Visual-Text Compression","","Long-context reasoning for large language models is computationally costly because self-attention scales quadratically with token length. Visual-text compression mitigates this by converting text to images and leveraging vision-language models, often reducing token consumption. Existing methods compress uniformly, ignoring query relevance and risking precision loss. SEER learns to selectively scan visual content, retrieve only query-relevant text, and then perform precise text-based reasoning with adaptive tool invocation. Benchmarks show improved extraction accuracy and substantial prompt-token savings.",{"@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/seer-long-context-reasoning-via-selective-visual-text-compression/273436/",{"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/seer-long-context-reasoning-via-selective-visual-text-compression/273436.png","ImageObject",300,407,{"name":92,"@type":93},"Lucas Martin","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-10","2026-09-15",true,{"@type":102,"interactionType":103,"userInteractionCount":56},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"Why is long-context reasoning expensive for large language models?","Question",{"text":112,"@type":113},"Self-attention has quadratic complexity over text tokens, making long sequences computationally costly.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"What problem do existing visual-text compression methods have?",{"text":117,"@type":113},"They apply uniform compression regardless of query relevance, which can sacrifice precision when fine-grained details are needed.",{"name":119,"@type":110,"acceptedAnswer":120},"How does SEER improve long-context extraction without losing efficiency?",{"text":121,"@type":113},"SEER selectively scans for query-relevant images, retrieves text only where needed, and then reasons over visual context plus retrieved text, improving accuracy while retaining average prompt-token savings.","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},273436,1789471427,{"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":56,"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":129,"read_time":144,"preview_page_count":4},8796095360427,"https://ap-avatar.wpscdn.com/davatar_994ba38a5ba835b3df7d355c54d3ed8d","arXiv :2608 . 15962v1 [ cs .CL] 16 Aug 2026  \nSEER: Long-Context Reasoning via Selective Visual-Text Compression  \nJiawei Xu1,∗ Zhilin Zhai2,∗ Jinrui Fang1 Ruohan Xu3 Mingfei Lu4 Yi Zhang4 Guanchu Wang5,† Tianlong Chen2,† Ying Ding1,†  \n1School of Information, The University of Texas at Austin  \n2Department of Computer Science, The University of North Carolina at Chapel Hill  \n3Department of Engineering, University of Cambridge  \n4Faculty of Engineering and Information Technology, University of Technology Sydney  \n5Department of Computer Science, The University of North Carolina at Charlotte  \n∗ Equal contribution. †Equal senior authorship.  \n[jiaweixu@utexas.edu](jiaweixu@utexas.edu) [zzhilin@unc.edu](zzhilin@unc.edu) [jinrui@utexas.edu](jinrui@utexas.edu)  \n[rx232@cam.ac.uk](rx232@cam.ac.uk) [mingfei.lu@student.uts.edu.au](mingfei.lu@student.uts.edu.au) [Yi.Zhang@uts.edu.au](Yi.Zhang@uts.edu.au)[ ](Yi.Zhang@uts.edu.au)[guanchu.wang@charlotte.edu](guanchu.wang@charlotte.edu) [tianlong@cs.unc.edu](tianlong@cs.unc.edu) [ying.ding@ischool.utexas.edu](ying.ding@ischool.utexas.edu)  \nAbstract  \nLong-context reasoning remains computationally expensive for large language models due to the quadratic complexity of attention over text tokens.  \nVisual-text compression offers a promising alternative by rendering text into images and processing them with vision-language models, often reducing token usage. However, existing approaches apply uniform compression regardless of query relevance, potentially sacrificing precision where detailed extraction is required. We present SEER, a framework that learns to select query-relevant images through visual scanning and retrieve textual content only where needed, combining the efficiency of visual compression with the precision of text-based reasoning. Through supervised fine-tuning on tool-interaction trajectories, SEER learns adaptive tool invocation for selection and retrieval. Experiments on long-context benchmarks show that SEER improves extraction precision through selective text retrieval while retaining average prompt-token savings relative to full-text baselines. On LongBench, SEER achieves 51.11% average accuracy, outperforming the visual-text baseline Glyph-9B by 2.33 points and Qwen3-8B by 3.49 points.  \nCode can be accessed at [https://github.com/jiaweixu98/SEER](https://github.com/jiaweixu98/SEER).  \n1 Introduction  \nLong-context reasoning (Bai et al., 2024; Sui et al., 2025) is increasingly important for applications such as multi-turn dialogue (Maharana et al., 2024) and long-horizon tasks (Denget al., 2025; Merrill et al., 2026) . However, processing long sequences with large language models (LLMs) remains computationally expensive due to the quadratic complexity of self-attention (Vaswani et al., 2017) . This challenge has motivated various approaches, including extended positional encodings such as YaRN (Peng et al., 2024; Wu et al., 2024) and sparse or linear attention (Yang et al., 2024) .  \nRecently, visual-text compression has emerged as a promising paradigm for efficient longcontext processing (Cheng et al., 2026; Xing et al., 2025; Wei et al., 2025) . By rendering textual content into images and processing them with vision-language models (VLMs), this approach can provide substantial token compression (number of textual tokens to vision tokens, typically 3–4× in prior work) while maintaining competitive accuracy (Cheng et al., 2026; Luo et al., 2026a) . The key insight is that vision encoders can represent dense textual information more compactly than tokenized text, enabling faster prefilling.  \nDespite these advantages, existing visualtext compression methods apply uniform compression across the entire context, regardless of query relevance (Cheng et al., 2026) . This uniform treatment presentsa practical trade-off: aggressive compression improves efficiency but may lose finegrained details necessary for precise reasoning; conservative compression preserves details but s","cbCaigv9DLvP1mrM","https://ap.wps.com/l/cbCaigv9DLvP1mrM","pdf",539409,27,"English","# Abstract\n## Long-context reasoning cost\n## Visual-text compression approach\n## SEER framework and three-stage pipeline\n## Experimental results on LongBench\n## Contributions and related work","[{\"question\":\"Why is long-context reasoning expensive for large language models?\",\"answer\":\"Self-attention has quadratic complexity over text tokens, making long sequences computationally costly.\"},{\"question\":\"What problem do existing visual-text compression methods have?\",\"answer\":\"They apply uniform compression regardless of query relevance, which can sacrifice precision when fine-grained details are needed.\"},{\"question\":\"How does SEER improve long-context extraction without losing efficiency?\",\"answer\":\"SEER selectively scans for query-relevant images, retrieves text only where needed, and then reasons over visual context plus retrieved text, improving accuracy while retaining average prompt-token savings.\"}]","SEER - Long-Context Reasoning via Selective Visual-Text Compression | PDF",68]