INDEX

Explanations

numbers

The neuron fires on standalone multi‐digit numerals (large integer tokens) in the text.

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

 dikutip

-1.63

⨾

-1.40

it

-1.38

最近は

-1.32

のは

-1.28

-1.27

に

-1.27

<bos>

-1.27

<blockquote>

-1.25

 quoique

-1.24

POSITIVE LOGITS

the

1.43

pegno

1.39

1.38

に行った

1.36

 تلك

1.32

be

1.31

 invoke

1.29

へ向

1.25

 والذي

1.25

 griech

1.24

Activations Density 0.001%