INDEX

Explanations

historical authors and sources

The neuron fires on bits of text that look like scholarly citations or references—author names, work titles, dates/centuries, book or volume numbers, etc.

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

ૉ

-1.34

 strategy

-1.16

 currently

-1.16

-1.14

 test

-1.14

 grosir

-1.13

 email

-1.08

 analyze

-1.07

әм

-1.06

 tests

-1.05

POSITIVE LOGITS

 景色

1.33

of

1.27

 による

1.26

pherds

1.23

époque

1.18

Iw

1.16

記載

1.15

 narratives

1.14

 later

1.13

 цвят

1.13

Activations Density 0.036%