INDEX

Explanations

advice and warnings

The neuron fires on recommendation expressions, i.e. tokens like “recommended” (and variants such as “not recommended”).

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

how

-1.75

籥

-1.64

ப்பட்ட

-1.61

╽

-1.61

喾

-1.58

꜆

-1.56

いただき

-1.55

頂き

-1.53

隄

-1.53

we

-1.52

POSITIVE LOGITS

↵↵

1.86

 Reverso

1.60

aqui

1.35

永遠

1.34

 ligeramente

1.32

 vollständ

1.31

年底

1.27

.”

1.26

 実例

1.26

 обуча

1.25

Activations Density 0.024%