INDEX

Explanations

Windows, Kubernetes, tf, ng, ps

This neuron responds to the accented “é” in the word “Pokémon,” effectively detecting mentions of “Pokémon.”

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 with

-2.02

for

-1.73

頃

-1.61

faring

-1.55

urface

-1.55

cation

-1.54

ExtendWith

-1.53

xbe

-1.52

—“

-1.52

 else

-1.50

POSITIVE LOGITS

 refroid

1.52

紅色

1.41

官網

1.41

ままで

1.41

icône

1.40

انواع

1.40

 équilibre

1.40

を書いて

1.38

然後

1.38

Somewhat

1.36

Activations Density 0.175%