INDEX

Explanations

questions asking why

The neuron selectively fires on the word “Why” (as in “Why …?” headings), flagging question‐style section titles.

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

ebly

-1.24

久しぶりに

-1.20

 officio

-1.18

Spes

-1.16

brought

-1.13

AKB

-1.13

Proveedor

-1.11

 gewünschten

-1.10

昨日は

-1.10

Materiale

-1.10

POSITIVE LOGITS

you

2.11

it

1.75

 choose

1.59

we

1.57

and

1.41

YOU

1.41

 Should

1.39

It

1.36

 should

1.34

 Choose

1.27

Activations Density 0.042%