INDEX

Explanations

punctuation, particularly commas and periods

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

owi

-0.09

hangi

-0.07

ABI

-0.07

Ø§Ø¨

-0.07

rray

-0.06

gaard

-0.06

é¾Ħ

-0.06

ardon

-0.06

vÃŃ

-0.06

got

-0.06

POSITIVE LOGITS

etc

0.10

et

0.09

etc

0.07

 uninitialized

0.06

rupted

0.06

_ARG

0.06

actical

0.06

etes

0.05

 needle

0.05

Activations Density 0.002%