INDEX

Explanations

code structures and separators

The neuron strongly activates on the labels and pattern tokens used in switch‐ or case‐style statements (e.g. “case” patterns ending in “)” or the “default:” label).

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

 jednej

-1.45

 récemment

-1.30

对她

-1.27

 Entste

-1.27

 sogen

-1.24

 jemanden

-1.23

about

-1.22

Otra

-1.21

 วิ

-1.21

考えた

-1.20

POSITIVE LOGITS

to

1.88

 publicados

1.62

by

1.57

 from

1.55

at

1.55

 then

1.50

 perhaps

1.48

 skjø

1.45

 they

1.45

珦

1.44

Activations Density 0.001%