一作ACL 2026Poster· Natural Language Processing
River- LLM : Large Language Model Seamless Exit Based on KV Share
Yingtao Shen · An Zou
摘要
Large Language Models (LLMs) have demonstrated exceptional performance across diverse domains but are increasingly constrained by high inference latency. Early Exit has emerged as a promising solution to accelerate inference by dynamically bypassing redundant layers. However, in decoder-only arch...