INDEX

Explanations

positive quality and speed

The neuron is essentially inactive (it never fires on any token).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

two

-1.63

not

-1.52

 there

-1.39

all

-1.36

by

-1.36

 without

-1.34

for

-1.33

 still

-1.33

 There

-1.30

 just

-1.26

POSITIVE LOGITS

 FUCKING

1.37

!!!!!!!

1.20

 !!!!

1.15

!!!!!!!!!

1.14

 !!!!!

1.13

 evtl

1.12

ycznych

1.07

 '-';

1.07

↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.06

 אלה

1.05

Activations Density 0.043%