INDEX

Explanations

numeric and punctuation separators

The neuron fires on tokens that look like numeric literals (digits, decimal points, signs, or embedded underscores).

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

 ถ้า

-1.79

缡

-1.55

همچنین

-1.52

нование

-1.51

QUENCE

-1.51

 danne

-1.49

ALLOWED

-1.46

vri

-1.46

🕜

-1.46

滹

-1.45

POSITIVE LOGITS

its

2.02

 that

2.02

of

1.61

at

1.58

1.52

 until

1.41

any

1.40

1.34

 efforts

1.27

1.25

Activations Density 0.002%