INDEX

Explanations

material identification

The neuron is detecting capitalized multiword names and technical terms (proper nouns and specialized titles).

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

 veľmi

-1.50

嗵

-1.50

achter

-1.48

*}$

-1.42

;");

-1.42

*",

-1.42

rentiel

-1.42

 tjen

-1.41

utre

-1.41

sterne

-1.41

POSITIVE LOGITS

by

2.42

 even

2.36

 their

1.89

 before

1.89

 only

1.88

 always

1.84

as

1.82

Even

1.75

 first

1.73

 perhaps

1.72

Activations Density 0.144%