INDEX

Explanations

under a commander or title

The main thing this neuron does is detect occurrences of the word “under.”

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

 هناك

-1.45

חים

-1.43

[]=$

-1.37

[]:

-1.37

瘓

-1.36

StoryboardSegue

-1.34

 andere

-1.30

孥

-1.24

饽

-1.24

 there

-1.24

POSITIVE LOGITS

 with

1.88

 both

1.66

it

1.63

 either

1.63

wth

1.62

 that

1.61

 each

1.60

with

1.58

from

1.45

 Затем

1.45

Activations Density 0.037%