INDEX

Explanations

foreign languages

The neuron strongly activates on title-cased or otherwise highlighted content words (e.g. capitalized tokens, code keywords like “False”/“print,” proper names, foreign-script words, or standalone quoted terms).

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

🧇

-1.34



-1.30

 etek

-1.30

huo

-1.20

 here

-1.19

 this

-1.19

SNMP

-1.18

 rafra

-1.16

kiu

-1.16

<0xF4>

-1.15

POSITIVE LOGITS

Why

1.66

 have

1.41

From

1.34

If

1.33

 should

1.31

 might

1.27

 could

1.26

 time

1.26

The

1.23

 were

1.17

Activations Density 0.034%