INDEX

Explanations

date and time formats

The neuron consistently fires on numeric tokens or tokens containing sequences of digits (e.g. parameter IDs, timestamps, version numbers).

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

NÉ

-0.96

гих

-0.95

anego

-0.94

 autora

-0.93

áno

-0.92

 quince

-0.91

اية

-0.91

álních

-0.90

henden

-0.90

бурга

-0.89

POSITIVE LOGITS

case

0.86

 Association

0.85

 City

0.83

 günstig

0.81

 Step

0.80

 Street

0.80

 Treasurer

0.78

 case

0.78

 Spring

0.78

Cho

0.76

Activations Density 0.001%