INDEX

Explanations

patterns and their contexts

New Auto-Interp

Configuration

Prompts (Dashboard)

16,384 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Kir

-0.09

anc

-0.09

oren

-0.08

 poison

-0.08

eral

-0.08

oge

-0.08

ERAL

-0.08

pha

-0.08

oucher

-0.08

-ticket

-0.08

POSITIVE LOGITS

ed

0.20

 Pattern

0.17

 pattern

0.17

.Pattern

0.16

(pattern

0.15

-pattern

0.15

pattern

0.15

 patterns

0.14

 bald

0.14

 patter

0.14

Activations Density 0.031%