Asma Ghandeharioun (@ghandeharioun) 's Twitter Profile
Asma Ghandeharioun

@ghandeharioun

Sr. Research Scientist @GoogleDeepMind working on ML interpretability & human-centered AI, PhD from @MIT

ID: 708149090740281345

linkhttps://alum.mit.edu/www/asma_gh calendar_today11-03-2016 04:34:34

125 Tweet

2,2K Followers

522 Following

Asma Ghandeharioun (@ghandeharioun) 's Twitter Profile Photo
Asma Ghandeharioun
@ghandeharioun

8 months ago

🧵Can we “ask” an LLM to “translate” its own hidden representations into natural language? We propose 🩺Patchscopes, a new framework for decoding specific information from a representation by “patching” it into a separate inference pass, independently of its original context. 1/9

🧵Can we “ask” an LLM to “translate” its own hidden representations into natural language? We propose 🩺Patchscopes, a new framework for decoding specific information from a representation by “patching” it into a separate inference pass, independently of its original context. 1/9