INDEX

Explanations

securities or security guard

The neuron fires on occurrences of the “securit-” root—i.e. tokens like “security,” “securities,” “issuer,” “issue,” “offer,” “shares,” and related legal/financial terms around securities.

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

 erratic

-1.75

 erroneous

-1.59

 purposeful

-1.55

 treacherous

-1.54

 inadequate

-1.54

The

-1.51

There

-1.50

 cumbersome

-1.49

-1.47

cor

-1.45

POSITIVE LOGITS

 ativos

1.92

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.60

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.52

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.50

 thier

1.47

zmm

1.45

 verlangen

1.44

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.42

 adicionais

1.42

 such

1.41

Activations Density 0.011%