INDEX

Explanations

actions followed by context

This neuron activates on imperative/action verbs that appear in user instructions or directions.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

for

-1.80

what

-1.35

等级

-1.27

 quantitative

-1.27

 ні

-1.27

になると

-1.21

靼

-1.20

 $^{\

-1.18

POSITIVE LOGITS

the

1.76

 following

1.59

 either

1.58

 using

1.56

哙

1.54

衚

1.52

 Estou

1.45

 현재

1.36

 juſ

1.35

 utilising

1.32

Activations Density 0.081%