INDEX

Explanations

mathematical expression variables

The neuron fires on explicit exponentiation or power‐of patterns (e.g. “**” and numbers raised to powers).

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

になってしまう

-1.05

になれば

-0.97

arkhand

-0.96

 esprits

-0.93

為に

-0.91

of

-0.91

ГРА

-0.91

ほうがいい

-0.89

 lunghe

-0.89

之地

-0.88

POSITIVE LOGITS

henswürdigkeiten

1.22

']]

1.16

and

1.13

xique

1.05

velocityY

1.03

llons

1.00

⎪

0.96

atorio

0.96

 >=",

0.95

govia

0.94

Activations Density 0.031%