[Paper] DCO: Dynamic Cache Orchestration for LLM Accelerators through Predictive Management
The rapid adoption of large language models (LLMs) is pushing AI accelerators toward increasingly powerful and specialized designs. Instead of further complicat...