INDEX
Explanations
The neuron fires on time‐and‐sequence markers – words and phrases that signal temporal transitions (e.g. “while,” “as,” “over time,” “day,” “end”).
New Auto-Interp
Negative Logits
Ping
-0.08
AIDS
-0.07
فارسی
-0.06
نده
-0.06
Typ
-0.06
gnu
-0.06
工程
-0.06
INS
-0.06
obi
-0.06
_for
-0.06
POSITIVE LOGITS
ynchron
0.07
breached
0.07
merc
0.07
θε
0.07
souha
0.07
就
0.06
สล
0.06
меня
0.06
_TBL
0.06
cripp
0.06
Activations Density 0.044%