Balancing performance and innovation in AI-driven supply chains through temperature-scaled hallucination control
研究了生成式AI中温度缩放参数如何通过平衡准确性与创造力来优化供应链决策，通过九项实验评估不同温度设置对需求预测、库存管理等任务的影响，发现低温提升精度、高温促进创新、中温实现平衡。
This study investigates how temperature scaling in generative AI (GenAI) models optimises decision-making in supply chain management by balancing accuracy and creativity. It addresses the challenge of tailoring AI-generated outputs for diverse supply chain tasks, spanning demand forecasting, inventory management, strategic planning, and process innovation. The research conducts nine experiments across key areas, evaluating AI models at varying temperature settings (low, moderate, and high) to assess their impact on accuracy, feasibility, and innovation. Results show that lower temperatures enhance precision and reliability, supporting operational efficiency, while higher temperatures foster creativity and innovation, benefiting strategic applications. Moderate temperatures strike an effective balance, enhancing adaptability in dynamic environments. The study identifies temperature scaling as a critical mechanism for improving AI-driven supply chain strategies, enabling managers to fine-tune AI models according to specific objectives. It contributes to the growing literature on AI in supply chain management by offering a structured approach to maximise AI’s value in both operational and strategic decision-making.