INDEX

Explanations

action followed by goal

This neuron detects the action-oriented verbs and associated function words that appear in how-to, what-to, where-to, and similar instructional or call-to-action headlines.

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

he

-1.47

 montrer

-1.30

刄

-1.20

是在

-1.18

 composto

-1.13

 पहले

-1.13

を観

-1.13

 WHILE

-1.13

czyk

-1.11

 выглядит

-1.11

POSITIVE LOGITS

or

1.75

 with

1.64

any

1.52

 gydy

1.43

 enkelte

1.35

 when

1.34

ᴉ

1.34

danh

1.33

 strtotime

1.30

 mitään

1.27

Activations Density 0.302%