INDEX

Explanations

The neuron primarily responds to question‐mark punctuation (i.e. the ends of questions).

questions and pros

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

 this

-1.38

回戦

-1.14

 information

-1.12

not

-1.05

for

-1.00

 Sosial

-0.95

on

-0.94

querque

-0.94

of

-0.93

 trouverez

-0.93

POSITIVE LOGITS

</strong>

1.30

 verrou

1.27

preferably

1.22

Dalam

1.20

ேன்

1.18

 escaleras

1.17

のですか

1.17

κος

1.16

 kılıf

1.15

 cinturón

1.15

Activations Density 0.004%