INDEX

Explanations

The neuron consistently activates on the “else” keyword in code.

else control flow

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

of

-2.44

and

-1.96

 then

-1.73

 from

-1.60

in

-1.56

but

-1.55

to

-1.52

for

-1.52

 except

-1.50

 along

-1.46

POSITIVE LOGITS

It

1.50

ナイキ

1.49

哈哈哈

1.47

棐

1.47

Setelah

1.43

 なお

1.42

܇

1.39

 aislado

1.39

 apprécier

1.39

 sèche

1.38

Activations Density 0.001%