#llm serving

Eight Papers Target Transformer Efficiency, Optimization, and Robustness

New arXiv batch addresses loss landscape myths, multi-LLM coordination, sparse caching, and adaptive computation depth.