INDEX

Explanations

describing relationships and common knowledge

The neuron specializes in spotting common methodological or relational terms—brief action and linking words (e.g. “using,” “due,” “rather,” “provides,” “known,” “do”) that signal procedures or connections in technical text.

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

//

-1.22

-1.19

ch

-1.18

-1.16

-1.15

PS

-1.15

 ایس

-1.14

-1.13

 الذي

-1.12

POSITIVE LOGITS

 Posteriormente

1.67

of

1.56

 muitas

1.54

 descrito

1.51

 potom

1.48

 internacionais

1.47

 nacionais

1.41

 ratón

1.38

iets

1.38

 Também

1.38

Activations Density 0.125%