INDEX

Explanations

The neuron detects discourse‐marker phrases of the form “in … ways” or “in … sense” (e.g. “in many ways,” “in a sense,” “a lot of ways”).

various ways, respects, sense, extent

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

鯔

-1.33

 població

-1.31

实的

-1.31

ครื่อง

-1.28

saya

-1.25

 parachoque

-1.23

 verzi

-1.23

Ꮵ

-1.23

desliz

-1.23

㈠

-1.23

POSITIVE LOGITS

it

2.56

 even

1.45

It

1.32

1.27

ot

1.23

1.22

 MORE

1.18

็

1.16

«

1.15

 this

1.14

Activations Density 0.010%