INDEX

Explanations

web file paths

The neuron detects tokens that look like file‐ or directory‐name segments in URLs or paths (e.g. asset and resource folder names).

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

 that

-2.44

for

-1.70

 always

-1.53

 Saltar

-1.40

 again

-1.38

 매우

-1.37

 either

-1.37

 also

-1.35

egd

-1.35

too

-1.33

POSITIVE LOGITS

Pendidikan

1.56

arities

1.52

 minore

1.48

 بعد

1.45

antMatchers

1.43

خدمات

1.38

مواد

1.38

رنگ

1.36

 vokal

1.35

ኧ

1.34

Activations Density 0.040%