INDEX

Explanations

identify potential vulnerabilities

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

醪

1.77

 آورد

1.64

ž

1.64

jeta

1.61

voerd

1.60

 оборо

1.59

 дела

1.59

яз

1.56

ie

1.55

 względu

1.55

POSITIVE LOGITS

 identific

2.29

 clues

2.27

 occurrences

2.27

 Treasures

2.24

蹤

2.22

 Detection

2.19

missing

2.19

 pitfalls

2.18

 Detect

2.17

 culprit

2.16

Activations Density 0.525%