[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-86585-en":3,"doc-seo-86585-105":30,"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":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},86585,7971461741311,"Ophelia","https://ap-avatar.wpscdn.com/avatar/74000253aff267980c6?x-image-process=image/resize,m_fixed,w_180,h_180&k=1779345379180704826",8,"Research & Report","HiFi-LLP High-Fidelity, Low-Cost Latency Predictors with Confidence for Robust HW-NAS","With deep neural networks increasingly deployed on edge devices, hardware-aware optimization such as HW-aware compression and HW-NAS depends on fast, reliable latency feedback from target hardware. Hardware-in-the-loop measurements form a sequential bottleneck, while latency predictors face sample-hunger and misleading errors. HiFi-LLP introduces graph-attention-based low-cost high-fidelity latency prediction with an explicit confidence metric. It improves accuracy bounds and reaches near-perfect rank correlation across devices, enabling a hybrid NAS that routes low-confidence cases to HIL.","HiFi-LLP: High-Fidelity, Low-Cost Latency Predictors with Confidence for Robust HW-NAS  \nShambhavi Balamuthu Sampath 1 ,2† Behzad Shomali3† Nael Fasfous2 Moritz Thoma 1 ,2 Judeson Anthony Fernando 1 Lukas Frickenstein2 Pierpaolo Mori2 Manoj Rohit Vemparala2  \nAlexander Frickenstein2 Walter Stechele 1  \n† Equal Contribution 1Technical University of Munich 2BMW Group 3University of Bonn  \narXiv :2607 . 1 1746v 1 [ cs .LG] 13 Jul 2026  \nAbstract—With deep neural networks (DNNs) increasingly deployed on edge devices, hardware (HW)-aware optimization techniques—such as HW-aware compression and HW-aware neural architecture search (HW-NAS)—have become essential. These methods rely on real feedback from the target hardware to tailor DNN architectures for efficient deployment. While the search can be parallelized, latency measurements via hardwarein-the-loop (HIL) remain a bottleneck due to their sequential nature. Recent approaches use latency predictors to replace costly HIL feedback, but challenges persist: (1) platform-specific predictors often require tens of thousands of samples, and (2) inaccurate predictions can mislead the NAS process. To address this, we introduce HiFi-LLP, a high-fidelity, low-cost latency predictor based on graph attention networks, augmented with a confidence metric. HiFi-LLP outperforms prior platform-specific predictors by up to 9 percentage points (p.p.) in the 10% accuracy bound and achieves a Spearman’s rank correlation of up to 0.996 across six devices in the LatBench dataset. We further propose a hybrid NAS framework that routes low-confidence predictions to HIL, achieving up to 8.6× speedup compared to typical NAS while maintaining a competitive Pareto front. 1  \nIndex Terms—Hardware-aware NAS, Hybrid NAS, Latency predictor, Confidence-aware predictor, Edge deployment  \nI. INTRODUCTION Safety-critical domains, such as autonomous driving, robotics, and industrial control, require algorithms that ensure high performance at real-time speeds. As DNNs are increasingly applied in these areas, the focus has shifted from merely achieving high accuracy to balancing accuracy with low latency. This shift requires architecture customization for specific target devices to meet both objectives, turning the task into a complex search across a large design space. To address this, NAS has evolved from optimizing for accuracy to HWaware NAS (HW-NAS), which considers both accuracy and  \nhardware constraints such as memory footprint and latency.  \nConventional HW-NAS frameworks, including HW-aware compression search, often rely on proxy metrics like floating point operations per second (FLOPS) or build exhaustive lookup tables (LUTs) for the search space. The latter is accurate  \n© 2025 IEEE. Personal use of this material is permitted. Permission from IEEE must be obtained for all other uses, in any current or future media, including reprinting/republishing this material for advertising or promotional purposes, creating new collective works, for resale or redistribution to servers or lists, or reuse of any copyrighted component of this work in other works. Published in the Proceedings of the 2025 IEEE 38th International System-on-Chip Conference (SOCC) . DOI: 10 . 1109/SOCC66126 .2025 .11235466  \n1 Correspondence to: [shambhavi.balamuthu-sampath@tum.de](shambhavi.balamuthu-sampath@tum.de)  \nbut is infeasible for vast search spaces, while the former is inaccurate, as lower FLOPS do not necessarily translate to lower latencies [1] . Therefore, recent works, especially for rapid NAS, have increasingly adopted latency predictors [1], [2] . Developed as a software solution, latency predictors are parallelizable with compute and eliminate the bottleneck of slow and sequential HIL measurements. State-of-the-art (SOTA) latency predictors often encounter a trade-off between latency predictor building time and their prediction accuracy. It is intuitive that with large amounts of training data resulting from exhaustive measureme","cbCaijrxjXejrILG","https://ap.wps.com/l/cbCaijrxjXejrILG","pdf",4102795,2,1,7,"English","en",105,"# Introduction\n# Related Works","[{\"question\":\"Why do hardware-in-the-loop (HIL) latency measurements limit HW-NAS?\",\"answer\":\"HIL latency measurements are sequential, creating a bottleneck when NAS must evaluate many candidate architectures.\"},{\"question\":\"What problem does HiFi-LLP address compared with existing latency predictors?\",\"answer\":\"It tackles the high sample requirements of platform-specific predictors and reduces harm from inaccurate predictions by adding a confidence metric to guide NAS decisions.\"},{\"question\":\"How does the proposed hybrid NAS use HiFi-LLP’s confidence?\",\"answer\":\"Low-confidence latency predictions are routed to HIL, while high-confidence predictions steer the search, yielding substantial speedups while keeping a competitive Pareto front.\"}]",1784212787,18,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":28},"hifi-llp-high-fidelity-low-cost-latency-predictors-with-confidence-for-robust-hw-nas","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,47,50],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":20},"https://docshare.wps.com/document/","Document",{"item":48,"name":12,"@type":43,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":43,"position":52},"https://docshare.wps.com/document/hifi-llp-high-fidelity-low-cost-latency-predictors-with-confidence-for-robust-hw-nas/86585/",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","2026-07-24","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},"Why do hardware-in-the-loop (HIL) latency measurements limit HW-NAS?","Question",{"text":75,"@type":76},"HIL latency measurements are sequential, creating a bottleneck when NAS must evaluate many candidate architectures.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What problem does HiFi-LLP address compared with existing latency predictors?",{"text":80,"@type":76},"It tackles the high sample requirements of platform-specific predictors and reduces harm from inaccurate predictions by adding a confidence metric to guide NAS decisions.",{"name":82,"@type":73,"acceptedAnswer":83},"How does the proposed hybrid NAS use HiFi-LLP’s confidence?",{"text":84,"@type":76},"Low-confidence latency predictions are routed to HIL, while high-confidence predictions steer the search, yielding substantial speedups while keeping a competitive Pareto front.","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":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,119,122,127,130,134],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":22,"doc_module":4,"doc_module_name":46,"category_name":116,"show_sort_weight":117,"slug":118},"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":120,"slug":121},30,"research-report",{"id":123,"doc_module":4,"doc_module_name":46,"category_name":124,"show_sort_weight":125,"slug":126},9,"Religion & Spirituality",20,"religion-spirituality",{"id":125,"doc_module":4,"doc_module_name":46,"category_name":128,"show_sort_weight":125,"slug":129},"World Cup","world-cup",{"id":131,"doc_module":4,"doc_module_name":46,"category_name":132,"show_sort_weight":131,"slug":133},10,"Lifestyle","lifestyle",{"id":135,"doc_module":4,"doc_module_name":46,"category_name":136,"show_sort_weight":106,"slug":137},19,"General","general"]