[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-84738-en":3,"doc-seo-84738-105":29,"detail-sidebar-cat-0-en-105":90},{"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":4,"is_deleted":4,"is_public":20,"is_downloadable":20,"audit_status":20,"page_count":21,"language":22,"language_code":23,"site_id":24,"html_lang":23,"table_of_contents":25,"faqs":26,"seo_title":13,"seo_description":14,"update_tm":27,"read_time":28},84738,137441390410,"Hazel","https://ap-avatar.wpscdn.com/avatar/2000252f4ab5702993?_k=1776741390130283984",8,"Research & Report","Compressing the Validation Bottleneck An Agentic Self-Driving Lab for Scientific Discovery","Agentic AI-for-Science automates ideation, planning, and analysis, but experiment-based validation remains a physical bottleneck that can increase both the number of trials and the cost per trial. The document introduces an agentic self-driving lab approach that targets two bottlenecks within a single agent: a prior-aware agentic DOE loop to improve trials-to-target, and a cost-aware surrogate measurement loop to predict expensive, high-resolution outcomes from cheap, low-resolution signals while managing uncertainty-driven selection. Experiments are discussed in biology and materials.","Compressing the Validation Bottleneck: An Agentic Self-Driving Lab for Scientific Discovery  \nKyunghoon Hur 1 Chihun Lee 2  \narXiv :2607 .04508v 1 [ cs .AI ] 5 Jul 2026  \nAbstract  \nAgentic AI-for-Science can automate ideation, planning, and analysis, but final validation still depends on real experiments. A self-driving lab (SDL) can execute those experiments, yet the loop still has bottlenecks: the agent may spend too many rounds on low-value experiments, or each round may require a high-cost experiment. We target these two physical bottlenecks with one agent. First, a prior-aware agentic DOE loop uses domain knowledge and past results to propose feasible and informative next experiments, reducing trials-to-target. Second, a cost-aware surrogate agent predicts high-cost, high-resolution measurements from low-cost, low-resolution measurements. It chooses between a high-and a lowcost measurement based on the predicted uncertainty. We examine these directions in the biology and materials domains, respectively. Together, under a single agent, these components aim to accelerate the SDL loop by reducing both the number of loops and the cost per experiment.  \n1. Agents Automate Everything but the Experiment  \nAgentic AI-for-Science can automate much of the research workflow apart from real-world experiments, spanning ideation, planning, experimental design, analysis, and writing, as illustrated by the Virtual Lab, the AI Scientist, and Coscientist (Swanson et al., 2025 ; Lu et al., 2024 ; Yamada et al., 2025 ; Boiko et al., 2023) . However, lab validation remains a physical constraint: a hypothesis is confirmed  \n1AX Research Division, Korea Electronics Technology Institute, Republic of Korea 2Material Data Division, Korea Institute of Materials Science, Republic of Korea. Correspondence to: Kyunghoon Hur \u003C[kyunghoonhur@keti.re.kr](kyunghoonhur@keti.re.kr) >, Chihun Lee \u003Cchi[hunlee@kims.re.kr](hunlee@kims.re.kr) > .  \nAccepted at International Conference on Machine Learning, Seoul, South Korea, 2026, AI for Science Workshop. AI Scientist Competition. Copyright 2026 by the author(s) .  \nonly when it is tested in real-world experiments. To mitigate this, as those upstream stages accelerate, the self-driving lab (SDL) has emerged as a promising way to reduce the time and cost of validation by automating repeated experiments and measurements (MacLeod et al., 2020 ; Szymanski et al., 2023) .  \nSDLs mainly focus on automating experiment execution, but a human still decides whether the proposed design of experiments (DOE) is feasible, useful, and scientifically informative. For an agentic SDL, that decision must be made by the agent itself, not by human intervention. We tackle two bottlenecks in this agent–bench loop: first, whether the agent can propose the next DOE by leveraging domain knowledge, feedback, and bench-side feasibility; and second, whether the agent can account for the time and resources needed to execute and measure each proposed experiment.  \n2. Bottleneck 1: The Agentic DOE Loop  \nIn an SDL, the agent proposes a DOE, the lab executesit under physical constraints, and the returned results inform the agent’s next proposal. The bottleneck is that this agent–bench interaction cannot always produce better results. Without combining domain priors, experimental feedback, and feasibility checks, the loop can proceed unproductively, spending rounds on low-value experiments.  \nBayesian optimization (BO) formalizes sequential experimental design by updating a surrogate with each returned result and selecting the next DOE by balancing promising conditions against uncertain regions (Frazier, 2018 ; Balandat et al., 2020) . However, vanilla BO usually encodes expert knowledge and laboratory constraints only indirectly, for example through fixed bounds, hand-designed variables, or manually curated search spaces. This motivatesan agent-driven BO-DOE loop in which domain-specific expert knowledge, literature priors, and bench-side const","cbCaitCK4wQZFSc2","https://ap.wps.com/l/cbCaitCK4wQZFSc2","pdf",5853257,1,6,"English","en",105,"# Agents Automate Everything but the Experiment\n# Bottleneck 1: The Agentic DOE Loop\n## Reducing trials-to-target with prior-aware proposals\n# Bottleneck 2: The Agentic Measurement Loop","[{\"question\":\"What validation bottleneck does the approach aim to reduce in agentic AI-for-Science?\",\"answer\":\"The approach reduces physical validation bottlenecks that increase the number of low-value experiment rounds and the cost required for each experimental-measurement round.\"},{\"question\":\"How does the method address bottleneck 1 in the agent–bench interaction loop?\",\"answer\":\"It uses a prior-aware agentic DOE loop that leverages domain knowledge, past results, and bench-side feasibility to propose feasible and informative next experiments, reducing trials-to-target.\"},{\"question\":\"How does the method address bottleneck 2 related to measurement cost?\",\"answer\":\"It employs a cost-aware surrogate agent that predicts high-cost, high-resolution measurements from low-cost, low-resolution signals and selects between high- and low-cost measurements based on predicted uncertainty.\"}]",1784197971,15,{"code":4,"msg":30,"data":31},"ok",{"site_id":24,"language":23,"slug":32,"title":13,"keywords":33,"description":14,"schema_data":34,"social_meta":85,"head_meta":87,"extra_data":89,"updated_unix":27},"compressing-the-validation-bottleneck-an-agentic-self-driving-lab-for-scientific-discovery","",{"@graph":35,"@context":84},[36,53,67],{"@type":37,"itemListElement":38},"BreadcrumbList",[39,43,47,50],{"item":40,"name":41,"@type":42,"position":20},"https://docshare.wps.com","Home","ListItem",{"item":44,"name":45,"@type":42,"position":46},"https://docshare.wps.com/document/","Document",2,{"item":48,"name":12,"@type":42,"position":49},"https://docshare.wps.com/document/research-report/",3,{"item":51,"name":13,"@type":42,"position":52},"https://docshare.wps.com/document/compressing-the-validation-bottleneck-an-agentic-self-driving-lab-for-scientific-discovery/84738/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":61,"encodingFormat":60,"isAccessibleForFree":62,"interactionStatistic":63},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-16",true,{"@type":64,"interactionType":65,"userInteractionCount":4},"InteractionCounter",{"@type":66},"ViewAction",{"@type":68,"mainEntity":69},"FAQPage",[70,76,80],{"name":71,"@type":72,"acceptedAnswer":73},"What validation bottleneck does the approach aim to reduce in agentic AI-for-Science?","Question",{"text":74,"@type":75},"The approach reduces physical validation bottlenecks that increase the number of low-value experiment rounds and the cost required for each experimental-measurement round.","Answer",{"name":77,"@type":72,"acceptedAnswer":78},"How does the method address bottleneck 1 in the agent–bench interaction loop?",{"text":79,"@type":75},"It uses a prior-aware agentic DOE loop that leverages domain knowledge, past results, and bench-side feasibility to propose feasible and informative next experiments, reducing trials-to-target.",{"name":81,"@type":72,"acceptedAnswer":82},"How does the method address bottleneck 2 related to measurement cost?",{"text":83,"@type":75},"It employs a cost-aware surrogate agent that predicts high-cost, high-resolution measurements from low-cost, low-resolution signals and selects between high- and low-cost measurements based on predicted uncertainty.","https://schema.org",{"og:url":51,"og:type":86,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":88,"canonical":51},"index,follow",{"doc_id":7,"site_id":24},{"code":4,"msg":5,"data":91},[92,96,100,104,109,113,118,121,126,129,133],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":93,"show_sort_weight":94,"slug":95},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":97,"show_sort_weight":98,"slug":99},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":101,"show_sort_weight":102,"slug":103},"Exam",70,"exam",{"id":105,"doc_module":4,"doc_module_name":45,"category_name":106,"show_sort_weight":107,"slug":108},5,"Comic",60,"comic",{"id":21,"doc_module":4,"doc_module_name":45,"category_name":110,"show_sort_weight":111,"slug":112},"Technology",50,"technology",{"id":114,"doc_module":4,"doc_module_name":45,"category_name":115,"show_sort_weight":116,"slug":117},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":119,"slug":120},30,"research-report",{"id":122,"doc_module":4,"doc_module_name":45,"category_name":123,"show_sort_weight":124,"slug":125},9,"Religion & Spirituality",20,"religion-spirituality",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":127,"show_sort_weight":124,"slug":128},"World Cup","world-cup",{"id":130,"doc_module":4,"doc_module_name":45,"category_name":131,"show_sort_weight":130,"slug":132},10,"Lifestyle","lifestyle",{"id":134,"doc_module":4,"doc_module_name":45,"category_name":135,"show_sort_weight":105,"slug":136},19,"General","general"]