[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-327629-105":59,"doc-detail-327629-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","evaluating-generative-agents-with-actions-grounded-in-socially-distributed-task-environments-using-incognita","Evaluating Generative Agents with Actions Grounded in Socially Distributed Task Environments using Incognita","","Effective agency in social environments hinges on when an agent seeks knowledge, when it acts, and whether actions are justified by acquired information. Existing grounded benchmarks separate executable actions and verifiable outcomes, while social simulation environments emphasize rich language-based interaction. This work unifies both by defining socially distributed task environments, introducing Incognita to separate social interaction from grounded execution, and evaluating multiple models on stratified tasks.",{"@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/evaluating-generative-agents-with-actions-grounded-in-socially-distributed-task-environments-using-incognita/327629/",{"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/evaluating-generative-agents-with-actions-grounded-in-socially-distributed-task-environments-using-incognita/327629.png","ImageObject",300,407,{"name":92,"@type":93},"Aria","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-30","2026-09-21",true,{"@type":102,"interactionType":103,"userInteractionCount":34},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What defines a socially distributed task environment in this paper?","Question",{"text":112,"@type":113},"Task-relevant knowledge is partitioned across role-isolated participants, and consequential actions are only achievable through interaction with those participants.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does Incognita separate social interaction from grounded execution?",{"text":117,"@type":113},"Agents route messages to user or specialist entities, specialist entities mediate admissible operations, a deterministic sub-environment executes accepted operations on a canonical state, and an offline evaluator scores final outcomes using inherited task rewards.",{"name":119,"@type":110,"acceptedAnswer":120},"What overall performance trends do the evaluated models show?",{"text":121,"@type":113},"Success increases from 0% to 8.9% and 17.2%, while premature finalization drops from 100% to 87% and 58%. Better models elicit more hidden knowledge and attempt more grounded writes, but reliability remains low.","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},327629,1790606634,{"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":34,"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":144,"read_time":145,"preview_page_count":4},2336464648322,"https://ap-avatar.wpscdn.com/avatar/2200025388227c56fec?_k=1778556882303663488","arXiv :2607 .02975v 1 [ cs .AI] 3 Jul 2026  \nEvaluating Generative Agents with Actions Grounded in Socially Distributed Task Environments using Incognita  \nDan C. Hsu1,2 Luke Lu1 ∗  \n1RedMind Research, San Francisco, CA, USA  \n2National Taiwan University, Taipei, Taiwan {dan, [luke](luke}@redmindresearch.org)[}](luke}@redmindresearch.org)[@redmindresearch.org](luke}@redmindresearch.org)  \nAbstract  \nEffective agency in social environments depends on when an agent seeks knowledge, when it acts, and whether its actions are justified by the information it has acquired. Existing grounded benchmarks provide executable actions, persistent state, and verifiable outcomes, while social simulation environments provide rich interaction among language agents. We study an evaluation setting that brings these requirements into a single environment.  \nWe define socially distributed task environments as interactive environments in which task-relevant knowledge is partitioned across role-isolated participants, and consequential actions are reachable only through interaction with those participants. Under this view, communication functions as exploration over role-partitioned knowledge, and grounded action functions as exploitation over consequential environment state. We introduce Incognita, a Concordia-based environment framework that instantiates this setting. Incognita separates social interaction from grounded execution.  \nThe evaluated agent routes messages to a user or specialist entities; specialist entities mediate admissible operations; a deterministic sub-environment executes accepted operations over a canonical state; and an offline evaluator scores final outcomes using inherited task rewards. Incognita-Retail transforms tau-bench retail into a multi-entity environment while preserving its final-state reward semantics. We evaluate three generative agent models on 18 tasks stratified by social breadth, with 540 total trials. Model progress appears in both reward and behavior. Success rises from 0 percent to 8.9 percent and 17.2 percent, while premature finalization falls from 100 percent to 87 percent and 58 percent. Stronger models elicit more hidden knowledge, contact more entities, and attempt more grounded writes, yet reliability remains low. These findings suggest that socially distributed task environments expose agent behavior before reliable task success, including knowledge elicitation, source selection, grounded action attempts, and premature completion belief.  \n1 Introduction  \nHuman action in organizations and service workflows is often conditioned by the social distribution of knowledge, where no single participant holds all task-relevant information and effective action depends on identifying who can provide which part of it (SCH¨UTZ, 1946) . Formal accounts of distributed knowledge likewise treat communication as a mechanism that changes what agents can know individually and collectively (Halpern & Moses, 1990) . As generative agents move from isolated instruction following to interactive environments, this structure becomes a central measurement problem for agent evaluation (Zhou et al., 2024a; Trivedi et al., 2024; Lu et al., 2025; Yao et al., 2025) . Evaluating such agents requires accounting for the final operation and the knowledge acquisition process that makes the operation appropriate (Yao et al., 2025; Laban et al., 2026; Qian et al., 2024; Li et al., 2025) .  \n∗ Corresponding author.  \nRecent grounded benchmarks have made substantial progress by placing agents in web, app, tool, and customer-service environments with executable actions, persistent state, and verifiable outcomes (Zhou et al., 2024a; Trivedi et al., 2024; Lu et al., 2025; Yao et al., 2025) . In parallel, work on LLM-based social simulation and social intelligence has shown that language agents can sustain believable social behavior, open-ended social interaction, and multi-actor worlds mediated by natural language (Park et al., 2023; Zhou et a","cbCaiaPNziZgQnTw","https://ap.wps.com/l/cbCaiaPNziZgQnTw","pdf",627805,13,"English","# Abstract\n## Socially distributed task environments\n## Incognita evaluation framework\n## Experimental results and analysis\n# 1 Introduction\n## Motivation from grounded benchmarks and social simulation","[{\"question\":\"What defines a socially distributed task environment in this paper?\",\"answer\":\"Task-relevant knowledge is partitioned across role-isolated participants, and consequential actions are only achievable through interaction with those participants.\"},{\"question\":\"How does Incognita separate social interaction from grounded execution?\",\"answer\":\"Agents route messages to user or specialist entities, specialist entities mediate admissible operations, a deterministic sub-environment executes accepted operations on a canonical state, and an offline evaluator scores final outcomes using inherited task rewards.\"},{\"question\":\"What overall performance trends do the evaluated models show?\",\"answer\":\"Success increases from 0% to 8.9% and 17.2%, while premature finalization drops from 100% to 87% and 58%. Better models elicit more hidden knowledge and attempt more grounded writes, but reliability remains low.\"}]","Evaluating Generative Agents with Actions Grounded in Socially Distributed Task Environments using Incognita | PDF",1789977669,33]