Zach does LLM interpretability and software
- Iām working on independent mech interp research these days!
- My current focus is hiearchical representations for interpretability of transformers. I want to see how we can use structure of learned features to better interpret language models and biological models.
Contact Me:
My Work:
- Preliminary Results on Graph SAEs: Building dependence graphs off of SAEs
- Research Notes: Sequence to Read Models: Notes on host to build (and not build) sequence-to-read genomic models