INDEX

Explanations

number followed by words like "count"

This neuron detects structured numeric identifiers (e.g. numbered headings, case or proclamation numbers, patent or document codes).

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

縢

-1.55

がありますが

-1.55

(!__

-1.54

 necessário

-1.54

 único

-1.52

 różnych

-1.50

 (!__

-1.49

𖤣

-1.43

我也不

-1.42

POSITIVE LOGITS

if

1.72

 there

1.59

↵↵

1.59

 that

1.41

Noon

1.35

 voordat

1.34

If

1.33

 Если

1.28

 Recuperado

1.27

),),

1.23

Activations Density 0.073%