INDEX

Explanations

admission, test, selection, Q:, names

The main thing this neuron does is detect formal institutional or bureaucratic terminology (e.g., exam names, ordinances, committees, government bodies, and admission processes).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 quando

-1.50

 τοῦ

-1.40

 when

-1.39

 cuando

-1.34

餚

-1.31

ὁ

-1.29

 ἐκ

-1.26

our

-1.21

ὑ

-1.21

 النبات

-1.18

POSITIVE LOGITS

ּוֹ

1.93

ִּ

1.46

͚

1.34

ֶּ

1.26

 бампер

1.26

 espoir

1.25

for

1.23

辦法

1.23

Legături

1.23

 '\\;'

1.21

Activations Density 0.015%