作者位次 #2/5ICML 2026Spotlight
Trojan-Speak: Bypassing Constitutional Classifiers with No Jailbreak Tax via Adversarial Finetuning
Bilgehan Sel · Xuanli He · Alwin Peng · Ming Jin · Jerry Wei
首次获取该候选人的完整档案及 StartLux 评估共 20 积分；每位用户单独解锁，之后查看免费。主动刷新 20 积分。
实时查找电子邮箱，费用不超过20积分；未找到邮箱不扣费。
Bilgehan Sel · Xuanli He · Alwin Peng · Ming Jin · Jerry Wei
Joshua Ong Jun Leang · Zheng Zhao · Aryo Pradipta Gema · Sohee Yang · Wai-Chung Kwan · Xuanli He · Wenda Li · Pasquale Minervini · Eleonora Giunchiglia · Shay B Cohen
Best-of- n sampling improves the accuracy of large language models (LLMs) and large reasoning models (LRMs) by generating multiple candidate solutions and selecting the one with the highest reward. The key challenge for reasoning tasks is designing a scoring function that can identify correct rea...