INDEX
Explanations
news articles
This neuron selectively activates on vivid, visceral action words—especially bodily or animal‐related acts (e.g. enema, flushing, chewing cud, hanging)—i.e. graphic or “gross” imagery.
New Auto-Interp
Negative Logits
注
-0.07
edeki
-0.07
Dic
-0.06
완
-0.06
ovel
-0.06
mary
-0.06
배
-0.06
quelle
-0.06
"){↵-0.06
혹
-0.06
POSITIVE LOGITS
progress
0.06
.ncbi
0.06
NewUrlParser
0.06
urg
0.06
0.06
angles
0.06
ActivatedRoute
0.06
+++
0.06
puls
0.06
_seqs
0.06
Activations Density 0.073%