1 paper across 1 session
Neural scaling law in LLMs is explained through representation interference due to superposition