INDEX

Explanations

uphill battle, long way, fighting with

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

 protéger

-0.84

 hält

-0.82

ляются

-0.81

 conduire

-0.77

もしっかり

-0.76

"";

-0.75

TRACE

-0.73

ષ

-0.72

École

-0.71

 établir

-0.71

POSITIVE LOGITS

 uphill

4.47

 hill

1.92

 downhill

1.91

 climb

1.80

hill

1.53

 steep

1.53

 upstream

1.48

 cuesta

1.46

 mountain

1.45

 hills

1.38

Activations Density 0.023%