INDEX

Explanations

by increasing or improving

The main thing this neuron does is detect numeric quantities and measurements (numbers and percentages).

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

 that

-1.89

 their

-1.55

 this

-1.39

 juos

-1.34

彼らは

-1.33

 cierta

-1.27

 ",";

-1.27

ふと

-1.27

=>

-1.27

beria

-1.27

POSITIVE LOGITS

 Verbraucher

1.48

jamientos

1.47

 chances

1.37

 мужа

1.35

чехол

1.34

heits

1.34

镲

1.33

自分が

1.32

で購入

1.32

 специаль

1.31

Activations Density 0.160%