INDEX

Explanations

phrases indicating potential improvements or suggestions

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

prompt

-0.07

rada

-0.07

RIX

-0.06

 genu

-0.06

elan

-0.06

ritt

-0.06

elage

-0.06

vette

-0.06

eks

-0.06

POSITIVE LOGITS

 improvement

0.10

 improve

0.09

 Improvement

0.09

 improved

0.08

 Improve

0.08

 improving

0.08

jit

0.08

 improves

0.07

 better

0.07

better

0.07

Activations Density 0.062%