INDEX

Explanations

inside or core concept

The neuron detects words or phrases that metaphorically point to inner or hidden qualities—tokens like “beneath,” “inside,” “deep,” or “down.”

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

in

-1.88

on

-1.81

 since

-1.73

 with

-1.73

if

-1.48

 this

-1.42

}$

-1.39

 because

-1.35

and

-1.30

😃

-1.28

POSITIVE LOGITS

 éto

1.48

 brisa

1.44

 peluche

1.44

 moelle

1.41

 selva

1.41

女友

1.40

1.37

︲

1.36

 falda

1.36

 jaune

1.36

Activations Density 0.024%