INDEX

Explanations

questions and answers

The neuron activates on customer-service review passages that describe a person’s helpful actions and responsiveness (e.g., “answered all the questions,” “took care of everything,” “quickly and thoroughly answered,” etc.).

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

to

-1.82

in

-1.66

it

-1.64

at

-1.57

by

-1.45

 began

-1.22

 begin

-1.21

 şeyi

-1.21

 should

-1.20

 with

-1.19

POSITIVE LOGITS

︲

1.51

 negru

1.51

夘

1.49

︐

1.49

voegd

1.45

縢

1.41

campista

1.38

 możemy

1.38

亖

1.38

 jezt

1.37

Activations Density 0.025%