Research

• Efficient Training of Sparse Autoencoders for Large Language Models via Layer Groups

  Davide Ghilardi, Federico Belotti, Marco Molinari, Tao Ma, Matteo Palmonari

  Cuts the cost of training sparse autoencoders by sharing them across groups of layers.

  EMNLP · 2025

• h4rm3l: A Dynamic Benchmark of Composable Jailbreak Attacks for LLM Safety Assessment

  Moussa Doumbouya, Ananjan Nandi, Gabriel Poesia, Davide Ghilardi, Anna Goldie, Federico Bianchi, Dan Jurafsky, Christopher D. Manning

  A composable language for synthesizing and benchmarking jailbreak attacks.

  ICLR · 2025

• Accelerating Sparse Autoencoder Training via Layer-Wise Transfer Learning in Large Language Models

  Davide Ghilardi, Federico Belotti, Marco Molinari

  Warm-starts sparse autoencoder training from neighboring layers to speed up convergence.

  BlackboxNLP @ EMNLP · 2024