AI RESEARCH
Mix-Quant: Quantized Prefilling, Precise Decoding for Agentic LLMs
arXiv CS.CL
•
ArXi:2605.20315v1 Announce Type: new LLM agents have recently emerged as a powerful paradigm for solving complex tasks through planning, tool use, memory retrieval, and multi-step interaction. However, these agentic workflows often