INDEX

Explanations

is followed by description

The neuron fires on copular “to be” forms (e.g. is, are, ’s, ’m).

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

 would

-1.40

 began

-1.39

ITCHEN

-1.33

ɛ

-1.33

 should

-1.31

 instead

-1.30

ったのは

-1.30

</strong>

-1.28

なんだけど

-1.27

 vpon

-1.27

POSITIVE LOGITS

 intercambio

1.40

randa

1.36

 таких

1.36

⑩

1.35

gridad

1.35

ご馳走

1.35

 אנחנו

1.32

僕も

1.32

 Endlich

1.29

 telles

1.28

Activations Density 0.028%