INDEX

Explanations

list of items preceding "included"

The neuron is primarily detecting capitalized or named-entity tokens (proper nouns and similar).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 with

-1.55

and

-1.52

 просто

-1.27

 there

-1.27

on

-1.20

in

-1.18

 They

-1.16

женская

-1.13

it

-1.11

 или

-1.11

POSITIVE LOGITS

whose

1.61

nadequate

1.52

despite

1.48

cannot

1.48

ऽ

1.47

~!

1.47

comes

1.47

蒌

1.44

܊

1.44

pecific

1.42

Activations Density 0.091%