INDEX

Explanations

empty string

The neuron activates on mentions of an “empty string,” i.e. occurrences of the phrase “empty string” (or “empty” and “string” together).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 arī

-1.73

 marchandises

-1.56

 myös

-1.54

 também

-1.48

 także

-1.46

 parachoque

-1.38

 bahkan

-1.38

 UserDao

-1.35

 icônes

-1.35

髕

-1.34

POSITIVE LOGITS

to

2.80

 until

1.55

 that

1.53

memcmp

1.48

</

1.46

 most

1.45

 when

1.44

1.39

if

1.37

Activations Density 0.049%