一作ICLR 2026Poster
On the Predictive Power of Representation Dispersion in Language Models
Yanhong Li · Ming Li · Karen Livescu · Jiawei Zhou
可解释性OpenReview
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。
Yanhong Li · Ming Li · Karen Livescu · Jiawei Zhou
Yanhong Li · Songlin Yang · Shawn Tan · Mayank Mishra · Rameswar Panda · Jiawei Zhou · Yoon Kim
Ming Li · Yanhong Li · Ziyue Li · Tianyi Zhou
As the post-training of large language models (LLMs) advances from instruction-following to complex reasoning tasks, understanding how different data affect finetuning dynamics remains largely unexplored. In this paper, we present a spectral analysis of layer-wise gradients induced by low/high-qu...