INDEX

Explanations

what + subject pronoun

The neuron is primarily detecting occurrences of the question word “what.”

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

 that

-1.83

こちらも

-1.70

 with

-1.67

for

-1.55

 what

-1.44

 kuid

-1.43

вались

-1.42

折叠

-1.33

,"

-1.33

 falle

-1.32

POSITIVE LOGITS

```

1.32

僕は

1.32

咙

1.29

ınızı

1.29

使

1.27

℣

1.23

曾經

1.19

 them

1.18

larınız

1.18

癲

1.17

Activations Density 0.015%