The field of large language models (LLMs) is witnessing significant advancements in optimization and automation. Recent developments focus on improving the efficiency and effectiveness of LLMs in various tasks, such as prompt optimization, speech synthesis, and operations research. Notably, innovative approaches are being explored to align LLMs with human feedback and preferences, leading to more accurate and natural results. Furthermore, automation techniques are being developed to streamline the process of modeling and solving complex problems, reducing the need for human expertise and labeled data. These advancements have the potential to significantly impact the field, enabling more robust and scalable LLM applications. Noteworthy papers include:

• A Toolbox for Improving Evolutionary Prompt Search, which proposes several key improvements to evolutionary prompt optimization, enhancing both optimization quality and efficiency.
• SpeechJudge, which introduces a comprehensive suite for speech naturalness judgment, comprising a dataset, benchmark, and reward model, and demonstrates superior performance in aligning speech synthesis with human preferences.
• OR-R1, which presents a data-efficient training framework for automated optimization modeling and solving, achieving state-of-the-art performance with limited labeled data.
• LoopTool, which introduces a fully automated, model-aware data evolution framework that closes the data-training loop for robust LLM tool calls, significantly enhancing tool-use capabilities.
• AutoSynth, which automates workflow discovery and optimization for high-quality synthetic dataset generation via Monte Carlo Tree Search, offering a scalable and cost-effective method for subjective LLM tasks.