INDEX

Explanations

can't / can

This neuron selectively activates on the modal verb “can” (including its negated form “can’t”).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ViewController

0.64

 unseen

0.60

Instead

0.57

Non

0.55

 avoided

0.55

というのは

0.55

 এড়িয়ে

0.54

 einem

0.54

 anche

0.54

æn

0.54

POSITIVE LOGITS

only

1.38

 only

1.35

 только

1.33

 нельзя

1.23

 فقط

1.22

 ONLY

1.21

 Только

1.20

 толькі

1.20

 tylko

1.19

 Only

1.18

Activations Density 0.176%