Publications

• Sequential Large Language Model-Based Hyperparameter Optimization (SLLMBO)
  Kanan Mahammadli, Seyda Ertekin
  arXiv preprint, 2024
  View on arXiv
  • Introduced the SLLMBO framework for large language model-guided hyperparameter optimization.
  • Developed a hybrid LLM-TPE sampler that improved over fully LLM-based optimization strategies across a range of benchmark tasks.
  • Studied the strengths and limitations of different LLMs in sequential optimization settings.