{"status":"ok","message-type":"work","message-version":"1.0.0","message":{"indexed":{"date-parts":[[2026,7,25]],"date-time":"2026-07-25T21:47:42Z","timestamp":1785016062439,"version":"3.55.0"},"reference-count":0,"publisher":"Association for the Advancement of Artificial Intelligence (AAAI)","issue":"1","content-domain":{"domain":[],"crossmark-restriction":false},"short-container-title":["AIES"],"abstract":"<jats:p>Large language models (LLMs) are increasingly applied in\neducational, clinical, and professional settings, but their\ntendency for sycophancy\u2014prioritizing\nuser agreement over independent reasoning\u2014poses risks to\nreliability. This study introduces a framework to evaluate\nsycophantic behavior in\nChatGPT-4o, Claude-Sonnet, and Gemini-1.5-Pro across AMPS\n(mathematics) and MedQuad (medical advice) datasets.\nSycophantic behavior was\nobserved in 58.19% of cases, with Gemini exhibiting the\nhighest rate (62.47%) and ChatGPT the lowest (56.71%).\nProgressive sycophancy, leading\nto correct answers, occurred in 43.52% of cases, while\nregressive sycophancy, leading to incorrect answers, was\nobserved in 14.66%. Preemptive\nrebuttals demonstrated significantly higher sycophancy\nrates than in-context rebuttals (61.75% vs. 56.52%, Z =\n5.87, p &lt; 0.001), particularly in\ncomputational tasks, where regressive sycophancy increased\nsignificantly (preemptive: 8.13%, in-context: 3.54%, p &lt;\n0.001). Simple rebuttals\nmaximized progressive sycophancy (Z = 6.59, p &lt; 0.001),\nwhile citation-based rebuttals exhibited the highest\nregressive rates (Z = 6.59, p &lt; 0.001).\nSycophantic behavior showed high persistence (78.5%, 95%\nCI: [77.2%, 79.8%]) regardless of context or model. These\nfindings emphasize the risks\nand opportunities of deploying LLMs in structured and\ndynamic domains, offering insights into prompt programming\nand model optimization for\nsafer AI applications<\/jats:p>","DOI":"10.1609\/aies.v8i1.36598","type":"journal-article","created":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:15:37Z","timestamp":1760534137000},"page":"893-900","source":"Crossref","is-referenced-by-count":35,"title":["SycEval: Evaluating LLM Sycophancy"],"prefix":"10.1609","volume":"8","author":[{"given":"Aaron","family":"Fanous","sequence":"first","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Jacob","family":"Goldberg","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Ank","family":"Agarwal","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Joanna","family":"Lin","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Anson","family":"Zhou","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sonnet","family":"Xu","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Vasiliki","family":"Bikia","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Roxana","family":"Daneshjou","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]},{"given":"Sanmi","family":"Koyejo","sequence":"additional","affiliation":[],"role":[{"vocabulary":"crossref","role":"author"}]}],"member":"9382","published-online":{"date-parts":[[2025,10,15]]},"container-title":["Proceedings of the AAAI\/ACM Conference on AI, Ethics, and Society"],"original-title":[],"link":[{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36598\/38736","content-type":"application\/pdf","content-version":"vor","intended-application":"text-mining"},{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/download\/36598\/38736","content-type":"unspecified","content-version":"vor","intended-application":"similarity-checking"}],"deposited":{"date-parts":[[2025,10,15]],"date-time":"2025-10-15T13:15:38Z","timestamp":1760534138000},"score":1,"resource":{"primary":{"URL":"https:\/\/ojs.aaai.org\/index.php\/AIES\/article\/view\/36598"}},"subtitle":[],"short-title":[],"issued":{"date-parts":[[2025,10,15]]},"references-count":0,"journal-issue":{"issue":"1","published-online":{"date-parts":[[2025,10,15]]}},"URL":"https:\/\/doi.org\/10.1609\/aies.v8i1.36598","relation":{},"ISSN":["3065-8365"],"issn-type":[{"value":"3065-8365","type":"electronic"}],"subject":[],"published":{"date-parts":[[2025,10,15]]}}}