INDEX

Explanations

reporting verbs and sources

The neuron activates on verbs and phrases that attribute information to a source (e.g. “says,” “said,” “according to,” “announced,” “informed,” “asserts”).

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

 либо

-1.43

 When

-1.42

 когда

-1.29

 either

-1.22

When

-1.20

 like

-1.16

傑

-1.15

<bos>

-1.14

十

-1.13

be

-1.11

POSITIVE LOGITS

 célè

1.30

 transférez

1.30

 incrí

1.29

Precalculus

1.28

 THEATRE

1.27

 brune

1.27

 engel

1.27

their

1.26

ってもら

1.26

がよく

1.25

Activations Density 0.136%