INDEX

Explanations

more particularly

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

 however

-0.12

however

-0.10

orne

-0.08

orn

-0.08

 However

-0.08

mont

-0.08

ICI

-0.07

 sorte

-0.07

ane

-0.07

However

-0.07

POSITIVE LOGITS

 importantly

0.15

 later

0.13

 bonus

0.11

enci

0.11

 worse

0.10

bonus

0.10

 twist

0.10

 optionally

0.10

 harder

0.09

PD

0.09

Activations Density 0.162%