Core Insights - Cognizant's AI Lab has developed a novel method for fine-tuning large language models (LLMs) using evolution strategies (ES), which promises to reduce training costs and improve accuracy compared to traditional reinforcement learning (RL) methods [3][4]. - The lab has been granted two new U.S. patents, bringing its total to 61, reinforcing its leadership in AI innovations [5][6]. Group 1: AI Innovations - The new research titled "Evolution Strategies at Scale: LLM Fine-Tuning Beyond Reinforcement Learning" demonstrates the successful application of ES for fine-tuning LLMs with billions of parameters, marking a significant advancement over RL methods [3][4]. - The ES approach requires less training data and enhances the quality of AI outputs, addressing the limitations of RL, which can be expensive and difficult to scale [4][5]. Group 2: Patent Details - U.S. Patent No. 12,424,335 focuses on AI-based optimized decision-making for epidemiological modeling, utilizing neural networks to predict trends like COVID-19 by integrating LSTM models for case and intervention histories [6]. - U.S. Patent No. 12,406,188 describes a system for evolved data augmentation and selection, which employs population-based search to enhance model robustness and performance with limited datasets [6][7]. Group 3: Future Directions - The AI Lab aims to scale its ES fine-tuning method to optimize the largest available LLMs for various complex tasks, following a 10X speed-up achieved through infrastructure improvements [4][5]. - The lab's mission is to maximize human potential through Decision AI, which combines generative AI, multi-agent architecture, and deep learning to create advanced decision-making systems [8][9].