INDEX
Explanations
references to small size or smallness
New Auto-Interp
Negative Logits
bsolute
-0.17
esa
-0.16
iras
-0.15
иÑĩа
-0.14
anon
-0.14
_MP
-0.14
ansen
-0.14
ollapse
-0.13
assemble
-0.13
andest
-0.13
POSITIVE LOGITS
/small
0.24
-scale
0.23
/tiny
0.21
(er
0.19
ish
0.19
llll
0.18
-small
0.18
-sized
0.18
/big
0.16
ledge
0.16
Activations Density 0.040%