INDEX

Explanations

assessment of

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

/non

-0.12

allee

-0.12

 COMM

-0.12

 ATTR

-0.11

 Dorm

-0.11

ades

-0.11

ous

-0.11

enant

-0.11

æĦ

-0.11

erver

-0.11

POSITIVE LOGITS

/testing

0.16

 inspections

0.13

 inspection

0.12

 probe

0.11

 scrutiny

0.11

 inspect

0.11

 probes

0.11

 evalu

0.10

 conducted

0.10

 probing

0.10

Activations Density 0.373%