INDEX

Explanations

standard procedures

The neuron consistently lights up on the discourse marker “Typically,” i.e. it detects that specific adverb in technical text.

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

{'

-1.79

\&

-1.70

 katkıda

-1.67

 katkı

-1.63

ificance

-1.58

-1.54

 eruption

-1.52

\%

-1.52

iencies

-1.50

 Bestellung

-1.48

POSITIVE LOGITS

 інших

1.99

 pié

1.70

 neum

1.66



1.59

 العالمي

1.54

ことはありません

1.50

 louer

1.46

 artesanal

1.45

 people

1.45

 worked

1.45

Activations Density 0.041%