INDEX

Explanations

lists within brackets

The neuron consistently fires on the single‐quote/apostrophe token.

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

↵

-2.02

-1.93

-1.91

“

-1.86

-1.77

 både

-1.69

section

-1.64

 vilá

-1.64

‘

-1.55

“-

-1.54

POSITIVE LOGITS

to

2.23

after

1.78

<th>

1.77

1.73

何を

1.68

すすめ

1.61

 pracu

1.60

 with

1.60

 disregarding

1.56

 教室

1.55

Activations Density 0.028%