INDEX

Explanations

final performance on unseen data

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

 Risiko

1.46

 Akar

1.39

 Zanz

1.39

 Caspian

1.34

 Tanık

1.34

谨

1.33

 limiti

1.33

 степень

1.32

 риск

1.31

собенности

1.30

POSITIVE LOGITS

 fooled

1.01

 actually

1.00

 Houd

0.99

 obtain

0.99

ardt

0.98

 fooling

0.95

adays

0.94

 thawed

0.94

👈

0.93

 CTRL

0.89

Activations Density 0.045%