作者位次 #3/6ICLR 2026Poster· Social Aspects / Fairness, Equity, Justice and Safety
RepIt: Steering Language Models with Concept-Specific Refusal Vectors
Vincent Siu · Nathan Henry · Nicholas Crispino · Yang Liu · Dawn Song · Chenguang Wang
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。
Vincent Siu · Nathan Henry · Nicholas Crispino · Yang Liu · Dawn Song · Chenguang Wang
Jianhong Tu · Nicholas Crispino · Kyle Montgomery · Chenguang Wang · Dawn Song
Visual text compression is an emerging paradigm for rendering text as images for processing by vision-language models (VLMs), enabling higher information density per context token. However, the robustness of VLMs under dense, text-based visual inputs remains unevaluated. We introduce Fico, a benc...