INDEX

Explanations

led by or coached by

This neuron detects colon characters (":") that introduce code directives, parameters, or metadata in programming and documentation contexts.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ueur

-0.75

 miejs

-0.68

FEATURE

-0.68

ویژگی

-0.67

 neve

-0.66

taken

-0.66

stad

-0.66

 polyps

-0.66

 crayon

-0.66

 REMOVE

-0.66

POSITIVE LOGITS

 здоровья

0.66

 phosphorus

0.62

 from

0.61

 Phosphorus

0.60

 Wildlife

0.60

bouncycastle

0.60

 doctrines

0.59

 thinks

0.59

Contacts

0.59

 rowspan

0.59

Activations Density 0.058%