INDEX

Explanations

web development and programming terms

The neuron activates primarily on isolated title‐style or proper‐noun tokens—i.e. capitalized words or headings.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

<bos>

-1.75

 such

-1.70

 from

-1.59

्रेस

-1.43

by

-1.41

"$_

-1.38

Such

-1.38

 Commande

-1.32

 poud

-1.30

 SIEM

-1.30

POSITIVE LOGITS

begin

1.63

==""){

1.62

 your

1.59

场比赛

1.58

」

1.56

">

1.55

 quiero

1.54

低

1.53

use

1.52

iniz

1.51

Activations Density 0.163%