[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-85671-en":3,"doc-seo-85671-105":29,"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":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},85671,549758252649,"Ivy","https://ap-avatar.wpscdn.com/avatar/8000253669c5317157?_k=1778319167496531819",8,"Research & Report","Transfer Learning Across Policy Regimes in Adaptive Multi-Agent Systems","Policy models often treat the policy–outcome mapping as stable across institutional conditions, an assumption that can fail in adaptive socio-technical settings where regulations reshape incentives and agents respond strategically. The paper frames regime change as transfer learning in adaptive multi-agent systems, comparing a blank-slate learner with a transfer learner using structural inductive bias. Results from emissions-regulation experiments and ABM robustness tests show transfer can improve small-sample performance under affine monotone tax–emissions structure, while threshold breaks cause negative transfer and persistent high prediction error.","arXiv :2607 .09685v1 [ cs .MA] 15 Jun 2026  \nTransfer Learning Across Policy Regimes in Adaptive  \nMulti-Agent Systems  \nRoberto Garrone  \nOpen University of Cyprus  \n[roberto. garrone@st. ouc. ac. cy](roberto. garrone@st. ouc. ac. cy)  \nAbstract  \nPolicy models often assume that the relationship between a policy instrument and its outcome remains stable across institutional conditions. In adaptive socio-technical systems this assumption may fail: regulatory change can alter incentives, agents can respond strategically, and the mapping from policy variables to aggregate outcomes can change. This paper studies such regime change as a transfer-learning problem in adaptive multi-agent systems. A policy regime is represented as a learning problem induced by an observable input distribution and a target function mapping policy variables to outcomes. We compare a blank-slate learner that searches a flexible hypothesis class in the new regime with a transfer learner whose effective hypothesis class is restricted by structural knowledge from the previous regime. Transfer is beneficial when this restriction preserves the new target function while reducing effective complexity; it is harmful when the restriction excludes the new target and creates misspecification. A stylized emissions-regulation experimental environment and a dynamic ABM robustness experiment support the claim. When the target regime preserves an affine monotone tax–emissions relation, transfer improves empirical small-sample performance. When the target regime introduces a threshold break, the same transferred structure produces negative transfer: held-out error remains high, online prediction generates more mistakes, and repeated online streams show larger cumulative and final-window error under misspecification. The contribution is methodological: previous regulatory experience should be reused when it captures stable structural invariants, but treated cautiously when policy change alters the policy–outcome relationship.  \n1 Introduction  \nLearning in real-world policy environments often occurs under changing institutional, technological, and regulatory conditions. Regulators and analysts do not observe passive systems. They intervene in environments populated by heterogeneous agents who interpret rules, respond to incentives, adapt their behavior, and interact with one another. These responses can feed back into the environment and change the relationship between policy instruments and aggregate outcomes. Environmental regulation, energy policy, urban planning, epidemiological intervention, and financial supervision all exhibit this general structure: a policy is introduced, agents respond, the system changes, and subsequent policy decisions are made using data generated under previous interventions.  \nSuch environments are naturally described as complex adaptive systems. Aggregate patterns emerge from heterogeneous interacting components rather than from a single centralized mechanism. Agent-based models (ABMs) are important in this context because they represent heterogeneous agents, bounded rationality, local interaction, and endogenous feedback explicitly [4, 6 , 11 , 17] . They are therefore well suited for studying policy interventions in adaptive socio-technical systems.  \nClassical learning theory begins from different assumptions. In the Probably Approximately Correct (PAC) framework, examples are usually assumed to be drawn from a stable distribution  \nand the target concept is fixed [13, 18] . These assumptions make it possible to derive samplecomplexity guarantees. One can ask how many observations are required for a learner to identify a hypothesis with error at most ϵ and confidence at least 1 − δ . However, these assumptions become problematic when the environment changes in response to policy intervention.  \nPolicy learning often proceeds as if the policy–outcome relationship is stable. A regulator may learn that a higher carbon tax tends to reduce ","cbCaipcPCiHdbdBK","https://ap.wps.com/l/cbCaipcPCiHdbdBK","pdf",726181,1,17,"English","en",105,"# Introduction\n## Policy learning under changing regimes\n## Adaptive socio-technical systems and ABMs\n## Transfer learning formulation and learners\n## Key contribution and experimental evidence","[{\"question\":\"How does the paper define a policy regime and the learning problem it induces?\",\"answer\":\"A policy regime is represented as an observable input distribution together with a target function mapping policy variables to outcomes. Regime change therefore changes both the distribution and the induced target mapping for the learner.\"},{\"question\":\"What is the difference between the blank-slate learner and the transfer learner in this study?\",\"answer\":\"The blank-slate learner searches a flexible hypothesis class in the new regime. The transfer learner restricts its effective hypothesis class using structural knowledge from the previous regime, treating transfer as structural inductive bias rather than warm-starting.\"},{\"question\":\"When does transfer improve performance, and when does it lead to negative transfer?\",\"answer\":\"Transfer improves empirical small-sample performance when the new target regime preserves an affine monotone tax–emissions relation. It becomes harmful when the new regime introduces a threshold break, producing high held-out and online prediction errors under misspecification.\"}]",1784205508,43,{"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":86,"head_meta":88,"extra_data":90,"updated_unix":27},"transfer-learning-across-policy-regimes-in-adaptive-multi-agent-systems","",{"@graph":35,"@context":85},[36,53,68],{"@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/transfer-learning-across-policy-regimes-in-adaptive-multi-agent-systems/85671/",4,{"url":51,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":23,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":40,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-17","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},"How does the paper define a policy regime and the learning problem it induces?","Question",{"text":75,"@type":76},"A policy regime is represented as an observable input distribution together with a target function mapping policy variables to outcomes. Regime change therefore changes both the distribution and the induced target mapping for the learner.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"What is the difference between the blank-slate learner and the transfer learner in this study?",{"text":80,"@type":76},"The blank-slate learner searches a flexible hypothesis class in the new regime. The transfer learner restricts its effective hypothesis class using structural knowledge from the previous regime, treating transfer as structural inductive bias rather than warm-starting.",{"name":82,"@type":73,"acceptedAnswer":83},"When does transfer improve performance, and when does it lead to negative transfer?",{"text":84,"@type":76},"Transfer improves empirical small-sample performance when the new target regime preserves an affine monotone tax–emissions relation. It becomes harmful when the new regime introduces a threshold break, producing high held-out and online prediction errors under misspecification.","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":24},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":20,"doc_module":4,"doc_module_name":45,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":46,"doc_module":4,"doc_module_name":45,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":52,"doc_module":4,"doc_module_name":45,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":45,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":45,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":45,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":45,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":45,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":45,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":45,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":45,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]