Skip to content
conftrace_
2025 ICML ICML 2025

OrthoRank: Token Selection via Sink Token Orthogonality for Efficient LLM inference

Abstract

No abstract available. No abstract was captured for this paper when its proceedings were scraped - most often an older edition published without one online. How the corpus is built.

Authors

Seungjun Shin, Jaehoon Oh, Dokwan Oh

Topics

Artificial Intelligence > Core AI > Interpretability Machine Learning > Core Methods > Representation Learning Artificial Intelligence > Core AI > Large Language Models

Keywords

hidden state token importance cosine similarity layer depth orthogonality sink token
Download PDF