[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86402-en":3,"doc-seo-86402-105":30,"detail-sidebar-cat-0-en-105":95},{"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":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},86402,1649267921044,"Ava Thompson","https://us-avatar.wpscdn.com/avatar/1800007509477c92dfb?_k=1782875107921204101",8,"Research & Report","2.5-D Decomposition for LLM-Based Spatial Construction","Autonomous systems that construct structures from natural-language instructions require dependable spatial reasoning, yet large language models often introduce systematic coordinate errors in three-dimensional block placements. The work proposes a neuro-symbolic pipeline using 2.5-D decomposition: an LLM plans only on the horizontal (2D) plane while a deterministic executor computes all vertical placements from column occupancy. Experiments on Build What I Mean (160 rounds) report 94.6% mean structural accuracy, surpassing GPT-4o and competing systems, with validated generalization to IGLU collaborative building tasks.","2.5-D Decomposition for LLM-Based Spatial Construction  \nPaul Whitten  \nRockwell Automation Mayfield Heights, OH, USA  0000-0002-7787-473X  \nLi-Jen Chen  \nRockwell Automation Mayfield Heights, OH, USA  0009-0000-8811-646X  \nSharath Baddam  \nRockwell Automation Mayfield Heights, OH, USA  0009-0008-8750-0548  \narXiv :2605 .07066v4 [ cs .AI] 11 Jul 2026  \nAbstract—Autonomous systems that build structures from natural-language instructions need reliable spatial reasoning, yet large language models (LLMs) make systematic coordinate errors when generating three-dimensional block placements. We present a neuro-symbolic pipeline based on 2.5-D decomposition: the LLM plans in the two-dimensional horizontal plane while a deterministic executor computes all vertical placements from column occupancy, eliminating an entire class of errors. On the Build What I Mean benchmark (160 rounds), GPT-4o-mini with this pipeline achieves 94.6% mean structural accuracy across 12 independent runs, within 3.0 percentage points of the 97.6% ceiling imposed by architect-agent errors that no builder-side improvement can address. This outperforms both GPT-4o at 90.3% and the best competing system at 76.3% . A controlled ablation confirms that 2.5-D decomposition is the dominant contributor, accounting for 28.7 percentage points of accuracy. The pipeline transfers directly to edge hardware: Nemotron-3 120B on an NVIDIA Jetson Thor AGX achieves 96.0% mean structural accuracy with the identical pipeline, slightly exceeding the cloud result. Expanding the system prompt by four targeted examples to exceed the model’s 8,320-token prefix cache page size, combined with low-effort reasoning, reduces mean perrequest latency by 3 × to 19.7 seconds at 95.6% accuracy. The underlying principle, removing deterministic dimensions from the LLM’s output space, applies to any autonomous construction or assembly task where gravity or other physical constraints fix one or more degrees of freedom. A transfer experiment on 500 IGLU collaborative building tasks confirms the effect generalizes beyond the primary benchmark.  \nIndex Terms—large language models, spatial reasoning, neurosymbolic systems, autonomous construction, 2.5-D decomposition  \nI. INTRODUCTION  \nAutonomous systems that construct physical structures from high-level instructions must solve two problems: understanding what to build and computing where to place each component. Large language models (LLMs) are effective at the first problem but unreliable at the second. When asked to produce three-dimensional coordinates, LLMs make systematic errors in vertical placement, off-by-one stacking, and duplicate positions [1], [2], consistent with LeCun’s observation that LLMs lack internal world models for enforcing physical constraints [3] .  \nWe observe that many construction domains exhibit a 2.5-D structure: one or more output dimensions are not free variables but deterministic functions of the others and the current state. In gravity-constrained block construction, the vertical coordinate of any new block is fully determined by the column occupancy below it. The LLM does not need to reason about this axis and in practice cannot do so reliably. Our approach separates the problem accordingly. The LLM produces a plan in the two-dimensional horizontal plane, specifying only (x, z) positions, colors, and action types. A deterministic spatial executor computes all vertical placement.  \nThis 2.5-D decomposition eliminates an entire class of coordinate errors. Fig. 1 shows a representative round in which the agent must recognize a T-shaped structure and extend it while preserving symmetry.  \nz  \ny  \nbuild  \ny  \nx z  \nx  \nFig. 1. A benchmark round requiring T-shape recognition. Instruction: “Keeping the T shape, extend the existing green structure by adding two green blocks to the longer base. Then add one purple block to each arm.” New blocks are marked with + .  \nThe system includes four additional components: a structure an","cbCaitxTiBCHQKmb","https://ap.wps.com/l/cbCaitxTiBCHQKmb","pdf",317066,3,1,6,"English","en",105,"# Introduction\n## Problem: LLM coordinate errors in 3D construction\n## Core idea: 2.5-D decomposition\n## System components and error ceiling\n# Related Work\n## LLM spatial reasoning\n## Plan-then-execute and neuro-symbolic decomposition","[{\"question\":\"What issue does the paper identify with LLMs in spatial construction tasks?\",\"answer\":\"When generating three-dimensional coordinates, LLMs make systematic mistakes in vertical placement, off-by-one stacking, and duplicate positions.\"},{\"question\":\"How does 2.5-D decomposition reduce coordinate errors?\",\"answer\":\"The LLM is restricted to planning in the 2D horizontal plane, while a deterministic executor computes vertical placements from column occupancy, eliminating unreliable vertical reasoning.\"},{\"question\":\"What performance improvements are reported on the Build What I Mean benchmark?\",\"answer\":\"With the proposed pipeline, GPT-4o-mini reaches 94.6% mean structural accuracy across 12 runs, outperforming GPT-4o (90.3%) and the best competing system (76.3%).\"},{\"question\":\"Does the approach generalize beyond the primary benchmark?\",\"answer\":\"A transfer experiment on 500 IGLU collaborative building tasks confirms the effect generalizes beyond the Build What I Mean benchmark.\"}]",1784211521,15,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":90,"head_meta":92,"extra_data":94,"updated_unix":28},"25-d-decomposition-for-llm-based-spatial-construction","",{"@graph":36,"@context":89},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":20},"https://docshare.wps.com/document/research-report/",{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/25-d-decomposition-for-llm-based-spatial-construction/86402/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-27","2026-07-16",true,{"@type":65,"interactionType":66,"userInteractionCount":20},"InteractionCounter",{"@type":67},"ViewAction",{"@type":69,"mainEntity":70},"FAQPage",[71,77,81,85],{"name":72,"@type":73,"acceptedAnswer":74},"What issue does the paper identify with LLMs in spatial construction tasks?","Question",{"text":75,"@type":76},"When generating three-dimensional coordinates, LLMs make systematic mistakes in vertical placement, off-by-one stacking, and duplicate positions.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does 2.5-D decomposition reduce coordinate errors?",{"text":80,"@type":76},"The LLM is restricted to planning in the 2D horizontal plane, while a deterministic executor computes vertical placements from column occupancy, eliminating unreliable vertical reasoning.",{"name":82,"@type":73,"acceptedAnswer":83},"What performance improvements are reported on the Build What I Mean benchmark?",{"text":84,"@type":76},"With the proposed pipeline, GPT-4o-mini reaches 94.6% mean structural accuracy across 12 runs, outperforming GPT-4o (90.3%) and the best competing system (76.3%).",{"name":86,"@type":73,"acceptedAnswer":87},"Does the approach generalize beyond the primary benchmark?",{"text":88,"@type":76},"A transfer experiment on 500 IGLU collaborative building tasks confirms the effect generalizes beyond the Build What I Mean benchmark.","https://schema.org",{"og:url":51,"og:type":91,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":93,"canonical":51},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":96},[97,101,105,109,114,118,123,126,131,134,138],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":106,"show_sort_weight":107,"slug":108},"Exam",70,"exam",{"id":110,"doc_module":4,"doc_module_name":46,"category_name":111,"show_sort_weight":112,"slug":113},5,"Comic",60,"comic",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":115,"show_sort_weight":116,"slug":117},"Technology",50,"technology",{"id":119,"doc_module":4,"doc_module_name":46,"category_name":120,"show_sort_weight":121,"slug":122},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":124,"slug":125},30,"research-report",{"id":127,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":129,"slug":130},9,"Religion & Spirituality",20,"religion-spirituality",{"id":129,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":129,"slug":133},"World Cup","world-cup",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":135,"slug":137},10,"Lifestyle","lifestyle",{"id":139,"doc_module":4,"doc_module_name":46,"category_name":140,"show_sort_weight":110,"slug":141},19,"General","general"]