LLMs with RAG enhance travel mode prediction accuracy

Researchers have developed a new framework for predicting travel mode choice using Large Language Models (LLMs) enhanced with Retrieval-Augmented Generation (RAG). The study evaluated four RAG strategies and three LLM architectures, including OpenAI's GPT-4o, o4-mini, and o3. Results showed that RAG significantly improved predictive accuracy, with the GPT-4o model combined with balanced retrieval and cross-encoder re-ranking achieving the highest accuracy of 80.8%. This approach also demonstrated superior zero-shot transfer abilities compared to traditional methods. AI

IMPACT Demonstrates LLMs can significantly improve predictive accuracy in specialized domains like transportation planning.

RANK_REASON Academic paper detailing a new methodology and experimental results. [lever_c_demoted from research: ic=1 ai=1.0]

Read on arXiv cs.AI →

• OpenAI GPT-4o
• Large Language Models
• o3
• o4-mini
• Puget Sound Regional Household Travel Survey
• Retrieval-Augmented Generation

• paper
• product

AI-generated summary · Google Gemini · from 1 sources. How we write summaries →