INDEX

Explanations

numbers following names/concepts

The neuron detects biblical scripture citations—i.e. book names followed by chapter and verse numbers (with colons and dashes).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

in

-1.75

 this

-1.18

of

-1.17

 since

-1.15

 even

-1.13

 would

-1.13

posals

-1.13

 jälkeen

-1.03

ったのは

-1.02

家は

-1.02

POSITIVE LOGITS

glected

1.20

rafo

1.20

Zé

1.11

ῖς

1.08

omio

1.08

ignen

1.06

 наиболее

1.05

ഛ

1.04

berday

1.04

 робота

1.03

Activations Density 0.004%