INDEX

Explanations

contains more

The neuron fires strongly on “action” words describing processes or instructions—i.e. verbs like mix, develop, cause, find out, etc.

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

 brawl

0.73

 fireball

0.71

 crap

0.71

 defrost

0.70

 Females

0.70

 havoc

0.69

 feces

0.69

 wacky

0.69

0.68

变成

0.67

POSITIVE LOGITS

 მხოლოდ

1.09

 sadece

0.96

だけでなく

0.96

only

0.95

 only

0.95

 hanya

0.93

เพียง

0.91

 tikai

0.90

 فقط

0.89

 yalnızca

0.87

Activations Density 0.839%