INDEX

Explanations

logical consequence or deduction

The neuron activates on numeric tokens (digits, measurements, and other numbers) wherever they appear in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

")

-1.80

')

-1.76

在

-1.71

ribute

-1.70

 peuvent

-1.66

 delicioso

-1.64

</h1>

-1.64

",

-1.58

 ที่

-1.58

 permita

-1.57

POSITIVE LOGITS

 that

2.52

 from

1.68

one

1.48

 since

1.44

 черво

1.35

 fact

1.34

 comes

1.34

 Beleuchtung

1.32

 their

1.31

 Они

1.31

Activations Density 0.018%