{"id":152,"date":"2026-01-12T07:26:37","date_gmt":"2026-01-12T07:26:37","guid":{"rendered":"https:\/\/wordpress.cs.vt.edu\/rylai\/?page_id=152"},"modified":"2026-08-26T02:01:18","modified_gmt":"2026-08-26T02:01:18","slug":"vexa","status":"publish","type":"page","link":"https:\/\/wordpress.cs.vt.edu\/traces\/vexa\/","title":{"rendered":"VEXA"},"content":{"rendered":"\n<p class=\"wp-block-paragraph\">Grounded but Misleading: Evaluating Semantic Alignment in AI-Generated Security Explanations<\/p>\n\n\n\n<figure class=\"wp-block-gallery has-nested-images columns-default is-cropped wp-block-gallery-1 is-layout-flex wp-block-gallery-is-layout-flex\">\n<figure class=\"wp-block-image size-large\"><img loading=\"lazy\" decoding=\"async\" width=\"1024\" height=\"861\" data-id=\"221\" src=\"https:\/\/wordpress.cs.vt.edu\/traces\/wp-content\/uploads\/sites\/257\/2026\/08\/grounding-illusion-1024x861.png\" alt=\"\" class=\"wp-image-221\" srcset=\"https:\/\/wordpress.cs.vt.edu\/traces\/wp-content\/uploads\/sites\/257\/2026\/08\/grounding-illusion-1024x861.png 1024w, https:\/\/wordpress.cs.vt.edu\/traces\/wp-content\/uploads\/sites\/257\/2026\/08\/grounding-illusion-300x252.png 300w, https:\/\/wordpress.cs.vt.edu\/traces\/wp-content\/uploads\/sites\/257\/2026\/08\/grounding-illusion-768x646.png 768w, https:\/\/wordpress.cs.vt.edu\/traces\/wp-content\/uploads\/sites\/257\/2026\/08\/grounding-illusion.png 1121w\" sizes=\"auto, (max-width: 1024px) 100vw, 1024px\" \/><\/figure>\n<\/figure>\n\n\n\n<p class=\"wp-block-paragraph\">Online scams increasingly leverage fluent and context-aware social-engineering strategies, creating growing demand for AI systems that explain why a message may be risky. However, explanations that cite detector-derived evidence may still semantically weaken or redirect the intended risk interpretation. We introduce VEXA (Verifying Semantic EXplanation Alignment), a controlled testbed for studying the gap between lexical grounding and semantic risk alignment in AI-generated scam-risk explanations. VEXA generates ungrounded, risk-aligned, and risk-diluting explanations by independently controlling evidence grounding and semantic framing. Across LLM-as-a-judge and human evaluations, explanations can remain comparatively grounded even when their interpretations weaken the detector\u2019s intended risk assessment. In human evaluation, risk-diluting XAI-grounded explanations retained comparatively elevated Perceived Evidence Grounding scores (3.66\u00b11.02) despite lower Helpfulness (3.00\u00b11.41) and Reasoning Support (3.14 \u00b1 1.05) scores. These findings provide controlled evidence of grounding-illusion effects and suggest that trustworthy explanation evaluation must verify not only whether evidence is cited, but also how it is interpreted.<\/p>\n\n\n\n<h2 class=\"wp-block-heading\">Publications<\/h2>\n\n\n\n<p class=\"wp-block-paragraph\">Heajun An, Connor Ng, Sandesh Sharma Dulal, Junghwan Kim, Jin-Hee Cho &#8220;Grounded but Misleading: Evaluating Semantic Alignment in AI-Generated Security Explanations.&#8221; In\u00a0<em>Proceedings of the 2026 AAAI\/ACM Conference on AI, Ethics, and Society<\/em> (AIES-26)<\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><a href=\"https:\/\/arxiv.org\/abs\/2602.05056\" data-type=\"link\" data-id=\"https:\/\/arxiv.org\/abs\/2602.05056\">Preprint<\/a><\/p>\n\n\n\n<p class=\"wp-block-paragraph\"><\/p>\n","protected":false},"excerpt":{"rendered":"<p>Grounded but Misleading: Evaluating Semantic Alignment in AI-Generated Security Explanations Online scams increasingly leverage fluent and context-aware social-engineering strategies, creating growing demand for AI systems that explain why a message may be risky. However, explanations that cite detector-derived evidence may still semantically weaken or redirect the intended risk interpretation. We introduce VEXA (Verifying Semantic EXplanation [&hellip;]<\/p>\n","protected":false},"author":511,"featured_media":0,"parent":0,"menu_order":0,"comment_status":"closed","ping_status":"closed","template":"","meta":{"footnotes":""},"class_list":["post-152","page","type-page","status-publish","hentry"],"_links":{"self":[{"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/pages\/152","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/pages"}],"about":[{"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/types\/page"}],"author":[{"embeddable":true,"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/users\/511"}],"replies":[{"embeddable":true,"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/comments?post=152"}],"version-history":[{"count":1,"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/pages\/152\/revisions"}],"predecessor-version":[{"id":222,"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/pages\/152\/revisions\/222"}],"wp:attachment":[{"href":"https:\/\/wordpress.cs.vt.edu\/traces\/wp-json\/wp\/v2\/media?parent=152"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}