INDEX

Explanations

number of

The neuron detects tokens containing digit sequences—that is, it fires on numbers embedded in identifiers or standalone numeric values.

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

and

-2.34

on

-1.90

 from

-1.80

in

-1.77

 with

-1.60

for

-1.48

at

-1.37

 should

-1.35

reddits

-1.31

があるので

-1.28

POSITIVE LOGITS

’

1.68

1.47

along

1.47

 lorsqu

1.42

坶

1.36

 ingresos

1.35

have

1.35

掲示

1.34

 appréci

1.33

 REALLY

1.32

Activations Density 0.020%