RKSC: Reasoning-Aware KV Cache Sharing and Confident Early Exit for Multi-Step LLM Inference 文章

ArXiv CS.CL2026-06-10NEWSen作者: Anirudh Sekar

RKSC: Reasoning-Aware KV Cache Sharing and Confident Early Exit for Multi-Step LLM Inference · 相关技术

相关技术