INDEX

Explanations

past and present roles

The neuron detects subject pronouns (e.g. “he,” “I”) or names together with auxiliary “be” forms and continuous/action verbs (the pattern he is/he had been working, Navin serves, I work, etc.).

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

 vijf

-1.51

 zeven

-1.39

ܞ

-1.36

 자신

-1.35

えます

-1.35

靦

-1.34

astră

-1.33

鲼

-1.32

になってしまう

-1.32

 usus

-1.30

POSITIVE LOGITS

as

2.33

an

1.78

 their

1.69

 providing

1.52

 provide

1.51

 additional

1.50

 provides

1.49

 making

1.46

↵↵

1.42

 started

1.41

Activations Density 0.037%