INDEX
Explanations
The neuron specifically detects the discourse adverb “also” (and its use as a sentence-opening transition).
New Auto-Interp
Negative Logits
ičky
-0.07
BED
-0.07
сред
-0.07
meaningful
-0.07
.tokens
-0.07
سپتامبر
-0.07
rollback
-0.07
SHR
-0.06
jed
-0.06
GetInt
-0.06
POSITIVE LOGITS
Also
0.09
Also
0.07
đồ
0.07
括
0.06
pollutants
0.06
交流
0.06
due
0.06
lee
0.06
hi
0.06
划
0.06
Activations Density 0.018%