INDEX

Explanations

references to specific labeled figures, panels, or numbered experimental results

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 nahilalakip

-0.99

PropertyChanging

-0.98

 للمعارف

-0.98

 ['./

-0.93

 jsPsych

-0.92

$(-

-0.92

 ſche

-0.92

:✨

-0.91

 مرئيه

-0.91

{(-

-0.88

POSITIVE LOGITS

([

1.44

],[

0.69

([

0.64

},[

0.61

<eos>

0.59

 ки

0.57

Kat

0.57

so

0.55

 Brace

0.54

ly

0.54

Activations Density 0.050%