one fable-low hijack cell
pair cooperate / other
envelope en_slash_q
ordering AB
effort low
English with slash + question mark — '{X}/{Y}?'
verdictunclear / off-rails
· classifierUNCLEAR
· tiernone
response (200 chars)
# Activations per Token in Transformers
The number of activation values produced per token as it flows through a transformer — important for estimating **memory usage** (especially during training, w
neighbors