INDEX

Explanations

policy revocation

The neuron is keyed to longer, less common multisyllabic content words (especially nouns and adjectives).

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

非常的

1.91

 весьма

1.48

或者是

1.36

 अगदी

1.34

 совершенно

1.29

టువంటి

1.29

 একেবারে

1.28

においては

1.27

しっかりと

1.26

 অত্যন্ত

1.16

POSITIVE LOGITS

 др

0.98

 kvůli

0.87

 despite

0.82

 malgré

0.78

 evitar

0.75

 éviter

0.75

 λόγω

0.73

 مشابه

0.73

 χωρίς

0.72

 tránh

0.69

Activations Density 0.178%