INDEX

Explanations

asking if correct

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

 recommandé

-1.09

aktionen

-1.03

unfortunately

-0.99

preferably

-0.91

 fortunately

-0.89

çuk

-0.89

 néhány

-0.89

fortunately

-0.85

 fubject

-0.85

başı

-0.85

POSITIVE LOGITS

 correct

4.47

correct

3.95

 Correct

3.81

Correct

3.78

 correctness

2.77

 incorrect

2.47

 CORRECT

2.47

CORRECT

2.39

 corrects

2.38

 Incorrect

2.20

Activations Density 0.109%