Softwares

• A toolkit for measuring context-mixing in text and speech Transformers [code]
• A disentanglement framework to seprate pre-trained speech representaions into textual and task-relevant acoustic features [code]

Datasets

• A French homophone dataset for context-aware ASR evaluation [data]

Tutorials

• “Interpretability Techniques for Speech Models” at Interspeech 2025 conference, in Rotterdam. [materials]
• Hands-on tutorial on “NLP & LLMs” for Dutch school students participating in the Olympiad in Artificial Intelligence (IOAI 2025).
• A series of short educational videos on how to open the Black Box of Large Language Models:
  • Why it is crucial to track how Transformers mix information? [YouTube]
  • How to best measure context-mixing in Transformers? [YouTube]
• “Transformer-specific Interpretability”, at EACL 2024 conference, in Malta. [materials]