About

This website holds the results and experiments done to achieve a mechanistic understanding of the Meta-(out-of-context)-learning phenomena that is observed in Large Language Models. This work was done as a part of AI Safety Camp, with Victor Levoso being the mentor. Here’s our team:

• Victor Levoso Fernandez
  • Research lead
  • figuring out planning and optimization in transformers
• Kunvar Thaman
  • website, social profiles
  • decoding transformations in transformers
• Leo Mckee-Reid
  • website
  • Team co-ordinator
  • ex-design engineer for orbital rockets
• Andrei Cristea
  • in-house Computer vision expert
  • Deep learning magician
• Florian van der Steen
  • experiments » theories
• Luan Fletcher
  • interested in technical work for AI alignment
• Nikita Menon
  • exploring ai safety niches