-
Restoring Identity: How Manifold Constraints Tame the Hyper-Connection Chaos
Why scaling Hyper-Connections breaks your model, and how a simple geometric trick fixes it.
-
Can Language Models Really ''Forget''? An Investigation into Machine Unlearning
A deep dive into whether making AI forget something actually deletes the knowledge, or just hides it
-
Peeking Inside the Black Box: My Deep Dive into Transformer Circuits
How Transformers actually work by breaking them down into simple, understandable circuits