INDEX

Explanations

numbers and times

The neuron activates on standalone digit tokens that begin numeric expressions—chiefly the initial digits of percentages, financial figures, scores, dates, and other numerical data in the text.

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

 choisi

-1.55

ofed

-1.52

 fournir

-1.49

eight

-1.48

amazing

-1.45

xiii

-1.45

 dieci

-1.43

跸

-1.41



-1.41

tais

-1.39

POSITIVE LOGITS

of

2.22

 only

2.17

 from

1.89

at

1.88

 seamlessly

1.79

 único

1.66

 before

1.64

↵↵

1.62

 through

1.61

 with

1.59

Activations Density 0.092%