INDEX

Explanations

specific phrases indicating responses or answers to questions

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

cerebras/SlimPajama-627B

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ibi

-0.07

laden

-0.07

igan

-0.07

imilar

-0.07

lig

-0.07

Ko

-0.06

ixin

-0.06

ums

-0.06

etsk

-0.06

agrid

-0.06

POSITIVE LOGITS

ing

0.08

UTILITY

0.07

ians

0.07

edImage

0.07

avin

0.06

ingu

0.06

edBy

0.06

ed

0.06

ov

0.06

Activations Density 0.010%