INDEX

Explanations

file path or name

The neuron fires on all-caps identifiers (often with underscores) that name constants, macros, environment variables, or other defined symbols in code.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

and

-3.64

or

-3.16

has

-2.50

 this

-2.33

as

-2.28

 before

-2.19

 these

-2.17

-2.11

the

-2.11

 becomes

-2.08

POSITIVE LOGITS

ݮ

2.45

 использование

2.41

赑

2.39

 eſſ

2.38

骉

2.34

 makam

2.30

 новая

2.30

 MUCH

2.27

assume

2.22

 diki

2.22

Activations Density 0.034%