INDEX

Explanations

introducing sections or locations

This neuron detects document-structure cues, especially section headings and introductory “this section…” phrases.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

it

-1.32

at

-1.16

and

-1.16

on

-1.07

 when

-0.99

 during

-0.96

 within

-0.92

an

-0.87

fällt

-0.86

Dorm

-0.85

POSITIVE LOGITS

 here

1.88

在这里

1.55

ここでは

1.54

 tässä

1.31

 здесь

1.30

 below

1.27

 anklicken

1.27

这里

1.26

here

1.23

 هنا

1.20

Activations Density 0.036%