INDEX

Explanations

okay

This neuron never activates—it’s essentially “dead” and doesn’t detect any pattern.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 Although

1.99

 During

1.79

 Throughout

1.73

 Though

1.71

 Shortly

1.70

 While

1.67

 Despite

1.66

 Through

1.66

 Besides

1.64

 Fortunately

1.63

POSITIVE LOGITS

Да

1.45

String

1.38

От

1.31

<blockquote>

1.27

def

1.27

in

1.27

name

1.26

get

1.25

And

1.24

No

1.24

Activations Density 0.320%