INDEX

Explanations

reporting findings and conclusions

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

We

-1.34

iramente

-1.33

its

-1.30

 this

-1.27

将其

-1.23

breviations

-1.21

gencias

-1.20

 betreft

-1.20

ઑ

-1.20

 इसे

-1.18

POSITIVE LOGITS

 there

1.95

 although

1.84

the

1.79

 хотя

1.68

 despite

1.58

although

1.57

 different

1.53

 even

1.48



1.46

 certain

1.41

Activations Density 0.154%