INDEX

Explanations

should therefore, should have, should be, should not

The neuron fires strongly on occurrences of the modal “should” (and related auxiliary constructions like “should be,” “should not,” etc.), marking prescriptive or normative statements.

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

 др

-1.45

 From

-1.42

-1.29

 still

-1.22

from

-1.21

şap

-1.21

 øns

-1.20

 Still

-1.20

 comunque

-1.17

 λό

-1.15

POSITIVE LOGITS

puntos

1.41

 setor

1.38

ฺ

1.38

荩

1.33

 يجب

1.33

個人的には

1.30

Ꮸ

1.23

これも

1.23

此人

1.22

痾

1.21

Activations Density 0.032%