INDEX

Explanations

and names

The neuron activates on numeric tokens—especially numbers in citations or reference parentheses.

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

-1.57

 ($("#

-1.52

 jöv

-1.51

年に

-1.46

 gelingt

-1.45

 guapo

-1.43

久しぶりに

-1.42

Nuestro

-1.41

目前

-1.39

maries

-1.39

POSITIVE LOGITS

of

2.25

 himself

1.67

 also

1.64

方がいい

1.55

 anaknya

1.46

SBATCH

1.46

 this

1.45

</i>

1.45

who

1.42

 sorento

1.42

Activations Density 0.007%