FlowDistill introduces a lightweight approach to traffic flow prediction by distilling knowledge from large language models, making smart transportation systems more accessible.

• Reduces computational overhead while maintaining high prediction accuracy
• Performs effectively even in data-scarce environments
• Leverages pre-trained LLM knowledge to enhance traffic prediction models
• Creates more practical solutions for real-world deployment in resource-constrained settings

Engineering Impact: This research bridges the gap between advanced AI capabilities and practical transportation applications, enabling more cities to implement intelligent traffic management without requiring massive computational resources or extensive historical data.

FlowDistill: Scalable Traffic Flow Prediction via Distillation from LLMs