INDEX

Explanations

The neuron selectively responds to capitalized proper‐name tokens, especially personal names and titles (e.g. people’s names, party names, official positions).

incumbent politicians

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

ENCIAS

-1.37

🍠

-1.34

Odpo

-1.33

 botanique

-1.31

ansvar

-1.30

蛀

-1.28

 αὐ

-1.27

 گیاه

-1.25

 międzynarod

-1.25

 メニュー

-1.24

POSITIVE LOGITS

1.30

on

1.27

 instead

1.26

ַּ

1.24

 party

1.20

ּוֹ

1.16

ִּ

1.15

 House

1.13

ָּ

1.13

versed

1.12

Activations Density 0.015%