INDEX

Explanations

references to roles and positions in various contexts

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

essen

-0.08

rael

-0.07

apon

-0.07

fold

-0.07

ree

-0.07

lad

-0.07

borg

-0.07

irs

-0.07

esse

-0.07

rame

-0.07

POSITIVE LOGITS

(Role

0.10

(role

0.08

.Role

0.08

playing

0.08

-playing

0.08

INGTON

0.08

.role

0.07

Rol

0.07

-role

0.07

Activations Density 0.020%