INDEX

Explanations

pronoun-verb constructions

The neuron strongly activates on personal references—pronouns (you, he, she, they) and proper names—i.e. tokens that refer to people.

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

Dlaczego

-1.13

of

-1.12

 freaking

-1.09

 göra

-1.09

Dzień

-1.08

 якщо

-1.06

 Direito

-1.05

Parece

-1.05

 Yoruba

-1.04

Curiosidades

-1.04

POSITIVE LOGITS

,"

1.34

۰۰

1.31

ってない

1.26

 zoude

1.23

'~

1.20

 añade

1.19

!')

1.18

 forbin

1.16

 other

1.16

了不少

1.15

Activations Density 0.035%