[{"data":1,"prerenderedAt":-1},["ShallowReactive",2],{"doc-detail-83633-en":3,"doc-seo-83633-105":30,"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":21,"is_downloadable":21,"audit_status":21,"page_count":22,"language":23,"language_code":24,"site_id":25,"html_lang":24,"table_of_contents":26,"faqs":27,"seo_title":13,"seo_description":14,"update_tm":28,"read_time":29},83633,1649267921044,"Ava Thompson","https://us-avatar.wpscdn.com/avatar/1800007509477c92dfb?_k=1782875107921204101",8,"Research & Report","ESC Emotional Self-Correction for Reliable Vision-Language Models","Vision-language models (VLMs) deliver strong performance on multimodal tasks, yet their reasoning can remain unreliable, often producing responses not supported by visual evidence. Self-correction approaches usually depend on post-training or carefully designed feedback, which increases computational cost. ESC (Emotional Self-Correction) revisits this problem by using emotion cues as a training-free trigger, enabling an external verifier to inject emotional feedback and encourage reflective revisions. Experiments across safety, hallucination, perception, and multimodal reasoning benchmarks improve reliability while preserving utility.","ESC : Emotional Self-Correction for Reliable Vision Language Models  \n2026  \nTien-Huy Nguyen 1 ,2 ,6⋆, Minh-Nhat Nguyen 1 ,3⋆, Nhat-Huy Nguyen4 ,6 , 11 ⋆ Hung-Viet Nguyen 1 ,4 ,6 , Huy Minh Nhat Nguyen 1 ,5 , Thanh-Huy Nguyen7 Cuong Tuan Nguyen5 , Hoang M. Le8 , Dat Nguyen9 , 10  \nPhat Kim Huynh 11 , Min Xu7 , 12 , Ulas Bagci 13†  \n1 GenAI4E Lab 2 University of Information Technology, Ho Chi Minh City, Vietnam  \n3 Universität Trier, Germany  \narXiv :2607 .02089v 1 [ cs .CV] 1 Jul  \n4 Ho Chi Minh University of Technology, Ho Chi Minh City, Vietnam  \n5 PAMI Lab, Vietnamese German University, Vietnam  \n6Vietnam National University, Ho Chi Minh City, Vietnam  \n7 Carnegie Mellon University, USA 8 Omoshiroi AI, USA  \n9 Harvard University, USA 10 Basis Research Institute  \n11 PASSIO Laboratory, North Carolina A&T State University, USA  \n12 Mohamed bin Zayed University of Artificial Intelligence, UAE  \n13 Northwestern University, USA  \n⋆ Equal contribution. † Corresponding author: [ulas.bagci@northwestern.edu](ulas.bagci@northwestern.edu)  \nFig. 1: Comparison of ESC against VLMs [50, 92] across diverse benchmarks.  \nAbstract. Vision-language models (VLMs) have achieved strong performance across diverse multimodal tasks, yet they remain vulnerable to unreliable reasoning. Existing self-correction methods mitigate these issues but typically rely on post-training or carefully engineered feedback, incurring high computational cost. In this work, we revisit this challenge through the lens of emotional cues, asking whether they can activate latent self-correction behaviors in VLMs without additional training. We  \n2 T.-H. Nguyen et al.  \nfind that emotional signals serve as an effective trigger for selfcorrection, encouraging more cautious and reflective reasoning. Motivated by this finding, we propose ESC (Emotional SelfCorrection), a training-free self-correction framework. ESC introducesan external verifier that detects potentially incorrect initial responsesand injects emotional feedback to encourage model to reflect, and produce a better revised response without additional training. Extensive experiments across safety, hallucination, vision-centric perception, and multimodal reasoning benchmarks show that ESC consistently improves reliability while preserving overall model utility. These results suggest that emotion can function not only as an ability to be recognized, but also as a practical control signal for scalable self-correction in VLMs. We therefore believe that ESC provides a strong foundation for anew reliable human-like, emotion-integrated research direction.  \nOur project is publicly available at [https://genai4e.github.io/ESC/](https://genai4e.github.io/ESC/) .  \nKeywords: VLMs · Self-Correction · Emotional Intelligence  \n1 Introduction  \n“AI models can have feelings too”  \nGeoffrey Hinton, 2024  \nThe progress of LLMs toward multimodal inputs [57,100] has enabled generalpurpose models for multimodal understanding via textual query. Among them, VLMs [52, 70, 87], which jointly process visual and textual input, demonstrate strong zero-shot capabilities in image search [59, 60] and VQA [61, 62] . Consequently, they are increasingly adopted in applications [24,25,80,91] . However, as VLMs are increasingly deployed in high-stakes domains such as healthcare [38] and security [94], a deeper understanding of their underlying behaviors is vital to identify and mitigate risks that may affect downstream applications. Despite their strong performance, existing studies [27,44] show that VLMs remain susceptible to hallucination, producing responses inconsistent with or unsupported by visual evidence. Prior work has sought to mitigate this issue through additional in-domain data [28,48,79], fine-tuning, or architectural modifications [41,45,51] . However, these approaches often require substantial computational resources.  \nTo motivate this limitation, we ask: How can VLMs correct their own mistakes at inference time without additional t","cbCaioOXrZyWkLaD","https://ap.wps.com/l/cbCaioOXrZyWkLaD","pdf",35053831,4,1,113,"English","en",105,"# Introduction\n## Motivation for inference-time self-correction\n## Emotion cues as a self-correction trigger","[{\"question\":\"What problem does ESC address for vision-language models?\",\"answer\":\"ESC addresses the vulnerability of VLMs to unreliable reasoning and hallucination, where responses may be inconsistent with or unsupported by visual evidence.\"},{\"question\":\"How does ESC perform self-correction without additional training?\",\"answer\":\"ESC uses an external verifier to detect potentially incorrect initial responses and injects emotional feedback to encourage the model to reflect and produce a revised answer.\"},{\"question\":\"What benefits does ESC show across benchmarks?\",\"answer\":\"ESC consistently improves reliability on safety, hallucination, vision-centric perception, and multimodal reasoning benchmarks while preserving overall model utility.\"}]",1784189402,285,{"code":4,"msg":31,"data":32},"ok",{"site_id":25,"language":24,"slug":33,"title":13,"keywords":34,"description":14,"schema_data":35,"social_meta":86,"head_meta":88,"extra_data":90,"updated_unix":28},"esc-emotional-self-correction-for-reliable-vision-language-models","",{"@graph":36,"@context":85},[37,53,68],{"@type":38,"itemListElement":39},"BreadcrumbList",[40,44,48,51],{"item":41,"name":42,"@type":43,"position":21},"https://docshare.wps.com","Home","ListItem",{"item":45,"name":46,"@type":43,"position":47},"https://docshare.wps.com/document/","Document",2,{"item":49,"name":12,"@type":43,"position":50},"https://docshare.wps.com/document/research-report/",3,{"item":52,"name":13,"@type":43,"position":20},"https://docshare.wps.com/document/esc-emotional-self-correction-for-reliable-vision-language-models/83633/",{"url":52,"name":13,"@type":54,"author":55,"headline":13,"publisher":57,"fileFormat":60,"inLanguage":24,"description":14,"dateModified":61,"datePublished":62,"encodingFormat":60,"isAccessibleForFree":63,"interactionStatistic":64},"DigitalDocument",{"name":9,"@type":56},"Person",{"url":41,"name":58,"@type":59},"DocShare","Organization","application/pdf","2026-07-23","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},"What problem does ESC address for vision-language models?","Question",{"text":75,"@type":76},"ESC addresses the vulnerability of VLMs to unreliable reasoning and hallucination, where responses may be inconsistent with or unsupported by visual evidence.","Answer",{"name":78,"@type":73,"acceptedAnswer":79},"How does ESC perform self-correction without additional training?",{"text":80,"@type":76},"ESC uses an external verifier to detect potentially incorrect initial responses and injects emotional feedback to encourage the model to reflect and produce a revised answer.",{"name":82,"@type":73,"acceptedAnswer":83},"What benefits does ESC show across benchmarks?",{"text":84,"@type":76},"ESC consistently improves reliability on safety, hallucination, vision-centric perception, and multimodal reasoning benchmarks while preserving overall model utility.","https://schema.org",{"og:url":52,"og:type":87,"og:title":13,"og:site_name":58,"og:description":14},"article",{"robots":89,"canonical":52},"index,follow",{"doc_id":7,"site_id":25},{"code":4,"msg":5,"data":92},[93,97,101,105,110,115,120,123,128,131,135],{"id":21,"doc_module":4,"doc_module_name":46,"category_name":94,"show_sort_weight":95,"slug":96},"Story & Novel",90,"story-novel",{"id":47,"doc_module":4,"doc_module_name":46,"category_name":98,"show_sort_weight":99,"slug":100},"Literature",80,"literature",{"id":20,"doc_module":4,"doc_module_name":46,"category_name":102,"show_sort_weight":103,"slug":104},"Exam",70,"exam",{"id":106,"doc_module":4,"doc_module_name":46,"category_name":107,"show_sort_weight":108,"slug":109},5,"Comic",60,"comic",{"id":111,"doc_module":4,"doc_module_name":46,"category_name":112,"show_sort_weight":113,"slug":114},6,"Technology",50,"technology",{"id":116,"doc_module":4,"doc_module_name":46,"category_name":117,"show_sort_weight":118,"slug":119},7,"Healthcare",40,"healthcare",{"id":11,"doc_module":4,"doc_module_name":46,"category_name":12,"show_sort_weight":121,"slug":122},30,"research-report",{"id":124,"doc_module":4,"doc_module_name":46,"category_name":125,"show_sort_weight":126,"slug":127},9,"Religion & Spirituality",20,"religion-spirituality",{"id":126,"doc_module":4,"doc_module_name":46,"category_name":129,"show_sort_weight":126,"slug":130},"World Cup","world-cup",{"id":132,"doc_module":4,"doc_module_name":46,"category_name":133,"show_sort_weight":132,"slug":134},10,"Lifestyle","lifestyle",{"id":136,"doc_module":4,"doc_module_name":46,"category_name":137,"show_sort_weight":106,"slug":138},19,"General","general"]