INDEX
Explanations
technical documents
This neuron detects the auxiliary “be” (and its inflected/modal forms like is, are, been, can, used to, etc.) as used in formal, patent‐style passive constructions.
New Auto-Interp
Negative Logits
manually
-0.07
بخشی
-0.06
been
-0.06
_parts
-0.06
녕
-0.06
范围
-0.06
отли
-0.06
fact
-0.06
cnt
-0.06
Seit
-0.06
POSITIVE LOGITS
mixture
0.08
cleaner
0.07
.savetxt
0.06
,module
0.06
座
0.06
ea
0.06
Sync
0.06
posta
0.06
cplusplus
0.06
afl
0.06
Activations Density 0.122%