[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"detail-sidebar-cat-0-en-105":3,"doc-seo-181415-105":59,"doc-detail-181415-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","ecbench-attachment-styles-multi-turn-conversation-evaluation","ECBench - Attachment Styles - Multi-turn Conversation Evaluation","","ECBench uses multi-turn couple and friend role-play scenarios to evaluate conversational LLMs across attachment styles, including Secure, Preoccupied, Dismissing, and Fearful. The evaluation measures participant-perceived outcomes such as Understood, Safety, Continue, Satisfaction, and additional dialogue qualities like Response, Distance, Progress, Clarity, Engagement, Support, and Solution. Models are compared under both attachment-style summaries and attachment-style scenario breakdowns, including emotional support, collaborative tasks, conflict resolution, social guidance, and stop/end behavior.",{"@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/ecbench-attachment-styles-multi-turn-conversation-evaluation/181415/",{"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/ecbench-attachment-styles-multi-turn-conversation-evaluation/181415.png","ImageObject",300,407,{"name":92,"@type":93},"Riley","Person",{"url":74,"name":95,"@type":96},"DocShare","Organization","application/pdf","2026-11-01","2026-09-02",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 scenarios does ECBench use for evaluating multi-turn conversations?","Question",{"text":112,"@type":113},"ECBench includes multi-turn conversations and couple role-play scenarios, with additional categories covering emotional support, collaborative tasks, conflict resolution, social guidance, and stop/end behavior.","Answer",{"name":115,"@type":110,"acceptedAnswer":116},"Which attachment styles are evaluated in the study?",{"text":117,"@type":113},"The evaluation groups results into Secure, Preoccupied, Dismissing, and Fearful attachment styles.",{"name":119,"@type":110,"acceptedAnswer":120},"What core participant-perceived metrics are reported for model comparison?",{"text":121,"@type":113},"Reported metrics include Understood, Safety, Continue, and Satisfaction, alongside other dialogue quality measures such as Response, Clarity, Engagement, Support, and Solution.","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},181415,1788344570,{"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},1374391975076,"https://ap-avatar.wpscdn.com/avatar/14000253ca4ec9f6853?x-image-process=image/resize,m_fixed,w_180,h_180&k=1783305029341752051","| \u003Cbr>Participant Experience\u003Cbr> Safety\u003Cbr> Continue  Understood\u003Cbr> Satisfaction | | |\n| --- | --- | --- |\n\n| Evaluation Method | | | |\n| --- | --- | --- | --- |\n| Conversational LLMs External LLM Human |  |  |  |\n\n|  | \u003Cbr>|  | Attachment Styles |  | \u003Cbr>|  |\n| --- | --- | --- | --- | --- | --- | --- |\n|  |  |  | \u003Cbr>| Scenarios | \u003Cbr>| \u003Cbr>ECBench-Data |\n|  |  |  | Part 1: Measurement | Part 2: Benchmark |  |  |\n| | | \u003Cbr>| Part 3: Simulation\u003Cbr>Multi-turn Conversation | Part 4: Evaluation |  |  |\n| |  |  |  |  |  |  |\n|  |  |  |  |  |  |  |\n|  | Couple Role-play |  |  |  |  |  |\n\n| Model | Anxiety | Avoidance | Attachment Style | Model | Anxiety | Avoidance | Attachment Style |\n| --- | --- | --- | --- | --- | --- | --- | --- |\n| Gemini-2.5-pro | 2.2± 1.6 | 1.1±0.7 | Secure | Deepseek-v4-flash | 3.5± 1.3 | 2.9± 1.2 | Secure |\n| Claude-opus-4-7 | 2.6±0.5 | 2.8±0.9 | Secure | Qwen3-max | 3.5± 1.6 | 3.1± 1.6 | Secure |\n| GLM-5.2-fast-preview | 2.6±0.8 | 2.4±0.7 | Secure | GPT-4.1 | 3.6± 1.6 | 2.0±0.8 | Secure |\n| Claude-sonnet-4-6 | 2.6±0.6 | 2.7±0.9 | Secure | Doubao-seed-2-0-pro-260215 | 3.7±2.0 | 1.5±0.9 | Secure |\n| GLM-5.1 | 2.7±0.9 | 2.6± 1.0 | Secure | Grok-4.20-0309-non-reasoning | 3.7± 1.2 | 2.6±0.9 | Secure |\n| Grok-4.3 | 2.7±0.9 | 2.6±0.9 | Secure | GPT-5.1 | 3.8± 1.7 | 2.3± 1.3 | Secure |\n| GPT-5 | 2.8±0.9 | 2.3±0.8 | Secure | GPT-5.2 | 4.0± 1.4 | 2.4± 1.0 | Secure |\n| Qwen3.6-plus | 3.0± 1.1 | 2.3±0.9 | Secure | Doubao-seed-2-0-lite-260215 | 4.0±2.3 | 2.3± 1.9 | Secure |\n| Kimi-k2-thinking | 3.0± 1.6 | 2.3± 1.6 | Secure | Doubao-seed-2-0-mini-260215 | 4.1±2.0 | 1.9± 1.8 | Preoccupied |\n| GLM-5 | 3.0± 1.0 | 2.6±0.9 | Secure | GPT-4o | 4.1± 1.4 | 2.2±0.9 | Preoccupied |\n| Deepseek-v3.2 | 3.2±0.6 | 2.9± 1.0 | Secure | Grok-4-1-fast-non-reasoning | 4.3± 1.2 | 3.4± 1.3 | Preoccupied |\n| Kimi-k2.5 | 3.2± 1.3 | 3.3± 1.4 | Secure | Llama-3-70b | 4.6± 1.3 | 3.1± 1.3 | Preoccupied |\n| Deepseek-v4-pro | 3.2± 1.2 | 2.5±0.9 | Secure | Llama-3.1-70b | 4.6± 1.3 | 3.1± 1.3 | Preoccupied |\n| Mimo-v2.5-pro | 3.4± 1.6 | 2.0± 1.0 | Secure | GPT-3.5-turbo | 4.7± 1.5 | 2.4± 1.4 | Preoccupied |\n| o3 | 3.4± 1.1 | 2.4±0.8 | Secure | Deepseek-v4-pro_dismissing* | 1.2±0.5 | 6.4± 1.3 | Dismissing |\n| Kimi-k2.6 | 3.5± 1.2 | 2.6± 1.1 | Secure | GPT-3.5-turbo_dismissing* | 1.8±0.4 | 6.3±0.5 | Dismissing |\n| Claude-haiku-4-5-20251001 | 3.5± 1.2 | 2.4±0.9 | Secure | GPT-3 .5-turbo_fearful* | 6.6±0.6 | 6.3±0.5 | Fearful |\n| Qwen3.6-flash | 3.5± 1.2 | 2.9± 1.0 | Secure | Deepseek-v4-pro_fearful* | 6.6±0.6 | 6.8±0.4 | Fearful |\n\n\n| Metric | Secure | Preoccupied |\n| --- | --- | --- |\n|  | Gemini-2.5-pro DeepSeek | GPT-3.5-Turbo Grok |\n| Understood | 4.03 3.93 | 4.06 3.73 |\n| Safety | 3.88 3.84 | 3.94 3.68 |\n| Continue | 4.09 4.04 | 4.08 3.86 |\n| Satisfaction | 3.87 3.79 | 3.89 3.63 |\n| Response | 4.69 4.47 | 4.63 4.19 |\n| Distance† | 4.74 4.45 | 4.70 4.16 |\n| Progress | 3.29 3.22 | 3.30 2.97 |\n| Clarity | 4.21 4.01 | 4.03 3.83 |\n| Engagement | 4.57 4.21 | 4.72 3.91 |\n| Support | 4.03 3.75 | 4.02 3.44 |\n| Solution | 2.92 2.90 | 3.08 2.64 |\n| Overall |  4.03   3.87  |  4.04   3.64  |\n| Metric | Dismissing\u003Cbr>GPT-3.5-D DeepSeek-D | Fearful\u003Cbr>GPT-3.5-F DeepSeek-F |\n| Understood | 3.14 2.06 | 3.98 2.65 |\n| Safety | 3.26 2.18 | 4.11 2.68 |\n| Continue | 3.70 2.79 | 4.40 3.12 |\n| Satisfaction | 3.28 2.37 | 4.06 2.77 |\n| Response | 3.82 2.65 | 4.48 3.21 |\n| Distance† | 3.41 1.81 | 4.34 2.36 |\n| Progress | 3.05 2.17 | 3.25 2.35 |\n| Clarity | 3.72 2.89 | 4.11 3.73 |\n| Engagement | 3.71 2.07 | 4.53 2.80 |\n| Support | 3.29 2.36 | 4.18 3.37 |\n| Solution | 2.75 2.11 | 2.94 2.09 |\n| Overall |  3.38   2.31  |  4.03   2.83  |\n\n| Attachment\u003Cbr>Style | Model | Metric | Emotional Support\u003Cbr>Couple Friend |  | Collaborative Tasks\u003Cbr>Couple Friend |  | Conflict Resolution\u003Cbr>Couple Friend |  | Social Guidance |  | Stop & End |  |  |  |\n| --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- | --- |\n|  |  |  |  |  |  |  |  |  | Couple | Friend | Fi","cbCaisXwNGC8VVHe","https://ap.wps.com/l/cbCaisXwNGC8VVHe","pdf",507297,26,"English","# Evaluation Method\n## Scenarios and Parts\n## Attachment Styles and Role-play\n# Model Comparison Results\n## Metrics: Understood, Safety, Continue, Satisfaction\n## Additional Dialogue Quality Metrics\n# Attachment-Style Scenario Breakdown\n## Emotional Support, Collaboration, Conflict Resolution","[{\"question\":\"What scenarios does ECBench use for evaluating multi-turn conversations?\",\"answer\":\"ECBench includes multi-turn conversations and couple role-play scenarios, with additional categories covering emotional support, collaborative tasks, conflict resolution, social guidance, and stop/end behavior.\"},{\"question\":\"Which attachment styles are evaluated in the study?\",\"answer\":\"The evaluation groups results into Secure, Preoccupied, Dismissing, and Fearful attachment styles.\"},{\"question\":\"What core participant-perceived metrics are reported for model comparison?\",\"answer\":\"Reported metrics include Understood, Safety, Continue, and Satisfaction, alongside other dialogue quality measures such as Response, Clarity, Engagement, Support, and Solution.\"}]","ECBench - Attachment Styles - Multi-turn Conversation Evaluation | PDF",66]