INDEX

Explanations

current tasks or next steps

This neuron fires on the discourse marker “Now,” i.e. when a new procedural step is being introduced.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

)}=

-1.56

↵↵↵

-1.51

跸

-1.44

舠

-1.41

 obdob

-1.38

("]");

-1.37

 Nicht

-1.33

to

-1.32

 všetkých

-1.32

"];

-1.30

POSITIVE LOGITS

are

1.59

Далее

1.53

 your

1.35

 recliner

1.32

鞄

1.29

Here

1.28

Like

1.26

 inet

1.26

优化

1.25

 auri

1.25

Activations Density 0.037%