-
RNNs Add Using a Helix Oct 3, 2025
The first part in a series applying interpretability methods to my bachelor's thesis.
-
Features and Logits: Bridging the Gap Sep 9, 2025
A preliminary investigation into connections between latent space steepness and downstream computation.
-
A brief look at how semantic entropy can be used to detect and reduce confabulations in LLMs.