INDEX

Explanations

clear or transparent

The neuron fires on occurrences of the adjective “clear” (i.e. transparent/see-through) modifying various 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

</b>

-1.56

需要

-1.43

 کنند

-1.42

 financi

-1.41

要是

-1.23

 Then

-1.23

 nächsten

-1.21

著名

-1.18

 graden

-1.16

POSITIVE LOGITS

\&

1.58

 clear

1.48

 прозра

1.42

クリア

1.42

 Transparent

1.40

 duide

1.39

 первом

1.39

 transparent

1.39

yeti

1.27

的基本

1.24

Activations Density 0.039%