作者位次 #3/5ICML 2026Poster· Deep Learning / Large Language Models
Beyond Perplexity: UTF-8 Validity in Byte-aware Language Models
Sangwhan Moon · Daisuke Oba · Youmi Ma · Tatsuya Hiraoka · Naoaki Okazaki
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。
Sangwhan Moon · Daisuke Oba · Youmi Ma · Tatsuya Hiraoka · Naoaki Okazaki
Youmi Ma · Naoaki Okazaki
Advances in mechanistic interpretability have identified special attention heads, known as retrieval heads, that are responsible for retrieving information from the context. However, the role of these retrieval heads in improving model performance remains unexplored. This work investigates whethe...