INDEX

Explanations

research methods and problems

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

 very

-0.96

brite

-0.85

ő

-0.84

 sehr

-0.84

 environnement

-0.83

似

-0.83

ということは

-0.82

 comparaison

-0.81

oners

-0.81

reshold

-0.79

POSITIVE LOGITS

 Meanwhile

0.91

 intrigu

0.90

 fanci

0.85

 Meine

0.85

扮演

0.84

 ACTUALLY

0.83

 retailing

0.82

 meanwhile

0.82

 feas

0.81

Meanwhile

0.81

Activations Density 0.065%