Summary

• Large Language Models (LLMs) face inefficiencies due to their substantial resource requirements.
• This survey categorizes optimization techniques into data-level, model-level, and system-level.
• Comparative experiments are conducted to quantify the benefits of these methods.
• Future research prospects are discussed, aiming to further enhance LLMs’ efficiency and application.

Opinion The deep dive into efficiency challenges provides critical insights for future innovations in AI, offering a roadmap for reducing computational overhead and accelerating LLM deployment.