INDEX

Explanations

high probability

The neuron detects numeric tokens or sequences (e.g. years, patent/application numbers, and other digit‐based values).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 tabloid

0.74



0.71

 frustrated

0.70

 arrogance

0.70

 stupidity

0.70

 stupid

0.69

 coordinating

0.68

soever

0.68

 ooit

0.67

 reconstru

0.66

POSITIVE LOGITS

㖑

0.63

тива

0.60

Съ

0.59

Ain

0.57

Bună

0.57

Hitting

0.57

ină

0.56

 Needle

0.56

ام

0.55

恋爱

0.55

Activations Density 0.000%