INDEX

Explanations

bigger number

This neuron fires on descriptive adjectives (i.e. modifiers) that qualify nouns.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

タイプの

0.77

版本的

0.65

couvrez

0.65

esempio

0.65

적으로

0.64

原因是

0.62

版的

0.62

ங்களில்

0.61

sächlich

0.61

migrations

0.60

POSITIVE LOGITS

 aldı

0.59

 staring

0.57

 ostensibly

0.57

 любы

0.56

 needlessly

0.56

 whatsoever

0.55

 coûte

0.55

 aussitôt

0.54

 enfe

0.54

<unused2199>

0.54

Activations Density 0.577%