LLM Circuit Analyses Are Consistent Across Training and Scale.

Curt Tigges Michael Hanna Qinan Yu Stella Biderman

Published in: CoRR (2024)

Keyphrases

neural network
scale space
training process
training set
supervised learning
high speed
training samples
evolvable hardware
databases
learning algorithm
feature selection
training data
power consumption
image structure
scale invariant
delay insensitive