LLMs and Applications

Created on February 20, 2025

2025   ·   llm   applications   recommendation   urban-intelligence

I explore how large language models and knowledge-enhanced learning can improve representation quality and decision intelligence in real-world urban and recommendation systems.

Key Works

  • LLMEmb (AAAI 2025): Using LLMs as high-quality embedding generators for sequential recommendation.
  • POI-Enhancer (AAAI 2025): Enhancing POI representation learning with LLM-driven semantic signals.
  • GARLIC (AAAI 2025): Integrating GPT-based reasoning with reinforcement learning for intelligent vehicle dispatching.

Research Focus

  • Build LLM-enhanced representations for downstream prediction and recommendation tasks.
  • Inject structured semantic knowledge into urban data mining pipelines.
  • Bridge language intelligence with operational decision-making systems.