INDEX

Explanations

dates and numbers

The neuron activates strongly on numeric tokens—especially date components, timestamps, and other digit sequences in update or version fields.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

}^{(\

-0.88

Apakah

-0.85

 своє

-0.84

írez

-0.83

 terminó

-0.82

ğını

-0.81

itability

-0.80

TOOLBAR

-0.79

 debió

-0.79

 knights

-0.79

POSITIVE LOGITS

 ChatGPT

0.84

 propon

0.84

に加

0.81

 entour

0.77

 prick

0.75

 statues

0.74

Pri

0.74

AI

0.74

Pri

0.73

Chat

0.73

Activations Density 0.012%