INDEX

Explanations

has been discussed or asked

The neuron detects metadiscourse cues that refer to earlier or external commentary (e.g. “earlier reviewer,” “as has been predicted/discussed,” “others pointed out,” “someone suggested”).

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

Then

-1.41

leneck

-1.27

 زیبا

-1.17

 glories

-1.12

↵↵

-1.11

wärts

-1.09

 imperialism

-1.09

Etter

-1.09

вшим

-1.08

 конечно

-1.06

POSITIVE LOGITS

previously

1.46

 said

1.42

phology

1.41



1.32

妧

1.30

𓆏

1.28

 laajenta

1.25

 others

1.22

Etimología

1.22

 financieras

1.21

Activations Density 0.043%