INDEX

Explanations

The neuron fires on high‐level topic nouns—i.e. the key subject words used as section or document headings (like “eruption,” “truck,” “neutron,” etc.).

specific word pairings

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 réinitial

-1.83

 déchir

-1.72

 réve

-1.69

 refroid

-1.68

 motoru

-1.57

 éte

-1.56

 coute

-1.56

 équilibr

-1.56

 nettoy

-1.55

殽

-1.55

POSITIVE LOGITS

as

2.02

1.91

in

1.77

1.63

1.62

1.57

in

1.56

1.53

for

1.52

Activations Density 0.083%