Researcher — paper.dinq.me

作者位次 #2/10ICLR 2026Poster· Deep Learning / Generative Models and Autoencoders

Next-ToBE: Probabilistic Next Token-Bag Exploitation for Activating Anticipatory Capacity in LLMs

Yihe Liu · Huibin Wang · Xianming Hu · Pinyi Zhang · Jiahao Xiong · Chenglin Wang · Nuoyi Chen · Hongbo Zhao · Jie Zhang · Kai Zhang

OpenReview

作者位次 #2/9ACL 2026Findings· Findings of ACL

Adaptive Zooming via Relevance-Informed Positional Resource Allocation for Training-free LLM Context Extension

Hongbo Zhao · Huibin Wang · Bin Tang · Xianming Hu · Yihong Huang · Yijun Shen · Nuoyi Chen · Ping Li · Kai Zhang

摘要

Large Language Models exhibit degraded performance when extrapolating beyond training context lengths. Existing training-free methods like positional reuse or interpolation can alleviate this issue in an efficient manner. However, these strategies are semantics-agnostic by only considering relati...

效率与系统长上下文与架构评测与基准OpenReview 项目页

Huibin Wang

获取完整研究档案

联系方式

2026 顶会论文

Next-ToBE: Probabilistic Next Token-Bag Exploitation for Activating Anticipatory Capacity in LLMs

Adaptive Zooming via Relevance-Informed Positional Resource Allocation for Training-free LLM Context Extension

常见合作者