[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-273425-105":59,"doc-detail-273425-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","zengen-social-mind-for-llms","ZenGen: Social Mind for LLMs","","Large language models moving from isolated task solving to long-term service in open environments need social intelligence—inferring mental states, tracking social relations, reasoning over norms, and adapting to situated context. The report introduces ZenGen (知境), an integrated framework covering social-intelligence measurement, internalizing social reasoning, and grounding it at deployment. It presents SoMBench, a comprehensive benchmark, and evaluates 20 LLMs, showing limited accuracy and low near-ceiling performance.",{"@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/zengen-social-mind-for-llms/273425/",{"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/zengen-social-mind-for-llms/273425.png","ImageObject",300,407,{"name":92,"@type":93},"Anda","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-10-29","2026-09-15",true,{"@type":102,"interactionType":103,"userInteractionCount":52},"InteractionCounter",{"@type":104},"ViewAction",{"@type":106,"mainEntity":107},"FAQPage",[108,114,118],{"name":109,"@type":110,"acceptedAnswer":111},"What social intelligence capabilities does the ZenGen framework target for LLMs?","Question",{"text":112,"@type":113},"It targets inferring mental states, tracking social relations, reasoning over norms, and adapting behavior under situated context.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"How does ZenGen measure social intelligence?",{"text":117,"@type":113},"It introduces SoMBench, a psychology-grounded benchmark defining social intelligence across primary and secondary dimensions and many task scenarios with expert-verified instances.",{"name":119,"@type":110,"acceptedAnswer":120},"What methods does ZenGen use to internalize social reasoning and improve performance?",{"text":121,"@type":113},"It develops a diagnosis-driven training recipe using supervised fine-tuning, on-policy distillation, and rubric-based reinforcement learning.","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},273425,1789471380,{"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":52,"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":129,"read_time":144,"preview_page_count":4},962075006959,"https://ap-avatar.wpscdn.com/avatar/e0002397efbe92a78e?_k=1776741047341049297","arXiv : 2607 .23740v2 [cs .CL] 2 1 Aug 2026  \nZenGen: Social Mind for LLMs  \nZenGen Team  \nAbstract  \nAs large language models move from isolated task solving toward long-term service inhuman environments, they require social intelligence: the ability to infer mental states, track social relations, reason over norms, and adapt behavior under situated context. This report presents ZenGen (知境), an integrated framework for measuring social intelligence, internalizing social reasoning, and grounding it at deployment time.  \nFor measurement, we introduce SoMBench, a psychology-grounded and comprehensive benchmark for social intelligence. It defines social intelligence as a structured capability space, covering 3 primary dimensions, 17 secondary dimensions, and 71 task paradigms. It further controls question format, narrative perspective, and context length across 284 shared scenarios and 3,481 expert-verified instances. Evaluation over 20 representative LLMs shows substantial remaining headroom: the best model reaches only 72.08% overall accuracy, and none of the 17 secondary dimensions reaches the 90% near-ceiling band.  \nFor internalization, we develop ZenGen, a diagnosis-driven training recipe using supervised fine-tuning, on-policy distillation, and rubric-based reinforcement learning. Across five social-cognition benchmarks, ZenGen consistently improves over its base models, with ZenGen-27B-Stage2 achieving the best average score among all compared models and ZenGen- 32B-Stage2 showing competitive performance against DeepSeek-V4-Pro. These results indicate that social reasoning can be strengthened as a stable model capability through diagnosisdriven staged training.  \nFor deployment-time grounding, we build Actio, a harness-controlled inference architecture that wraps a base LLM and routes four typed supports into reasoning: PRISM for procedural guidance, Starling for runtime mental-state representation, SAGE for reusable experience, and gated RAG for external social and normative knowledge. Across five base models and three social-cognition benchmarks, the full harness improves 14 of 15 modelbenchmark pairs and is best or tied for best in 8 pairs, showing that typed runtime support can systematically strengthen social reasoning at inference time. Module-level analyses attribute these gains to selective activation of complementary supports, establishing Actio asan effective path for grounding social reasoning at deployment time.  \nTogether, these results show that social mind LLMs require coordinated progress in evaluation, parametric internalization, and deployment-time grounding.  \nBenchmark performance of ZenGen (知境)-series and baselines  \n1 Introduction  \nModern large language models are increasingly used, and often envisioned, as a path toward general-purpose artificial intelligence. This vision has been supported by a clear capability progression. Early LLMs demonstrated strong language intelligence through fluent generation, knowledge recall, and instruction following; more recent models have moved toward task intelligence, showing increasingly strong abilities in reasoning, coding, planning, and tool use.  \nNow consider a stronger deployment scenario. An LLM is no longer a passive assistant that answers isolated questions, but a long-term collaborator embedded in open social environments. It must interact with users, groups, and other agents while reading social cues such as implicit goals, emotions, relationships, and norms. It must also act under incomplete information and evolving context. In such a setting, task intelligence alone is insufficient. A response can befactually correct but socially inappropriate; an action can be efficient but unsafe; a short-term objective can be satisfied while long-term trust is damaged. The central challenge therefore shifts from solving tasks to acting appropriately, reliably, and safely in social contexts.  \nThis reveals a missing layer of intelligence. Just as embodied agents mu","cbCaieBjS1PVpzc2","https://ap.wps.com/l/cbCaieBjS1PVpzc2","pdf",9127431,73,"English","# Introduction\n## Social intelligence as a missing layer\n## Related evaluation and enhancement work","[{\"question\":\"What social intelligence capabilities does the ZenGen framework target for LLMs?\",\"answer\":\"It targets inferring mental states, tracking social relations, reasoning over norms, and adapting behavior under situated context.\"},{\"question\":\"How does ZenGen measure social intelligence?\",\"answer\":\"It introduces SoMBench, a psychology-grounded benchmark defining social intelligence across primary and secondary dimensions and many task scenarios with expert-verified instances.\"},{\"question\":\"What methods does ZenGen use to internalize social reasoning and improve performance?\",\"answer\":\"It develops a diagnosis-driven training recipe using supervised fine-tuning, on-policy distillation, and rubric-based reinforcement learning.\"}]","ZenGen: Social Mind for LLMs | PDF",184]