INDEX

Explanations

not followed by contrast or exception

The neuron is detecting occurrences of the word “Not” (i.e. a sentence-initial or capitalized negation 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

茕

-3.08

༈

-2.72

呖

-2.66

 också

-2.56

 DELHI

-2.47

圌

-2.34

ης

-2.30

臯

-2.30

떳

-2.28

μος

-2.27

POSITIVE LOGITS

Not

3.03

3.00

 Importantly

2.69

瀦

2.50

 Whilst

2.50

 Conversely

2.47

.,

2.44

 Virtually

2.42

 They

2.42

);

2.42

Activations Density 0.009%