INDEX

Explanations

names and roles

The neuron fires on capitalized named‐entity tokens, especially personal names (surnames and given names) and similar proper nouns.

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

 which

-1.72

in

-1.55

and

-1.46

of

-1.43

at

-1.38

 what

-1.34

an

-1.30

 because

-1.29

 that

-1.25

 from

-1.24

POSITIVE LOGITS

 exuber

1.22

 reposer

1.20

 relentless

1.20

 remplacé

1.17

 ludicrous

1.15

 substituted

1.13

 évidemment

1.09

פיה

1.09

 horrid

1.09

 substitutes

1.09

Activations Density 0.008%