INDEX

Explanations

userusers of

This neuron fires on the central nouns of formal or technical multi-word phrases—i.e. the head nouns in titles, named works, and domain-specific 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

-1.78

or

-1.77

 other

-1.75

 after

-1.53

it

-1.52

 when

-1.43

in

-1.40

全新的

-1.34

 with

-1.27

 same

-1.27

POSITIVE LOGITS



1.87

濺

1.38

 soudain

1.36

竄

1.35

 konkre

1.35

 gena

1.32

Sì

1.32

who

1.32

 extraordinaire

1.30

 hulle

1.29

Activations Density 0.171%