INDEX

Explanations

scrutiny is, is television, forgiven for

The neuron consistently lights up on longer, less common or domain-specific nouns (especially multi-syllable or specialized terms).

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

 people

-1.48

 vanskelig

-1.45

 člove

-1.45

 ljudi

-1.40

 življen

-1.37

 ľudí

-1.37

 predstav

-1.35

 tendremos

-1.34

 dårlig

-1.30

 móg

-1.30

POSITIVE LOGITS

of

2.14

是

1.58

-------

1.56

tistic

1.44

and

1.44

the

1.43

--

1.41

 idyl

1.39

:""

1.39

----------

1.38

Activations Density 0.128%