I trained a sparse autoencoder on the middle layer residual stream of my genome language model and found human-interpretable latent features that consistently detect specific DNA motifs!
🧵1/8
Shae Mclaughlin
@shaemcl.bsky.social
29 followers 36 following 9 posts
Med student turned researcher studying gene regulation 🧬 | Currently MS (Health Data Science) @ UCSF 🌁 | Interested in: epigenome editing & deep learning 🧠