INDEX

Explanations

numbers and common words

This neuron detects long hexadecimal strings (e.g., revision or commit hashes).

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

 wszystkich

-1.33

his

-1.32

[]

-1.28

lampe

-1.27

ᕱ

-1.25

 vijf

-1.24

😶

-1.23

😕

-1.22

här

-1.18

 Địa

-1.18

POSITIVE LOGITS

 незавер

1.41

 that

1.35

losseum

1.30

 only

1.28

 which

1.13

to

1.12

 принад

1.12

 захворю

1.10

so

1.06

not

1.06

Activations Density 0.003%