INDEX

Explanations

numbers and punctuation

The neuron fires most strongly on very frequent, common tokens (e.g. conjunctions, prepositions, commas)—in other words, high‐frequency “stop” words and punctuation.

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

 these

-1.59

 this

-1.58

 both

-1.30

—“

-1.28

 various

-1.28

 after

-1.27

—”

-1.27

などなど

-1.20

two

-1.17

 many

-1.16

POSITIVE LOGITS

1.89

</strong>

1.63

</i>

1.55

</em>

1.41

</b>

1.38

").

1.22

 obligator

1.21

 innumer

1.20

[.

1.19

<tr>

1.16

Activations Density 0.059%