The explosive success of large language models (LLMs) has captured the attention of the deep learning research community, and their powerful capabilities have demonstrated remarkable performance in various fields, such as natural language processing and computer vision. For example, since the ChatGPT of OpenAI has been online, it has accumulated over 10 billion users. However, LLMs face significant efficiency challenges, especially when deploying large neural networks with billions of parameters on edge devices such as personal computers and smartphones, which is impractical. In addition, LLMs have shown strong reliability issues (e.g., adversarial vulnerability, biased predictions, hallucinations, etc.), further impeding their applications in practice. These challenges have provided a strong impetus for model compression/robustness research in the era of LLMs, leading to the development of resource-efficient and robustness-enhancement methods. The research theme has brought challenges and opportunities to the deep learning community, requiring experts in deep learning, natural language processing, and computer vision to adopt interdisciplinary approaches to uncover the underlying mechanisms and develop practical solutions. By studying and addressing the efficiency and reliability challenges associated with LLMs, we could better enable practitioners to construct practical LLMs across various domains. The workshop will bring together researchers and practitioners from the machine learning communities to explore the latest advances and challenges in building practical learning methods, with a special focus on the efficiency and reliability of LLMs. The workshop consists of invited talks by leading experts in the field and contributed talks and poster sessions featuring the latest research.