INDEX

Explanations

repetition or future events

The neuron detects words and punctuation that signal repetition or continuation, such as “again,” “another,” and ellipses.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

</strong>

-1.69

 FOURTH

-1.43

翃

-1.34

眼底

-1.34

Carrot

-1.34

 THIRD

-1.33

 levert

-1.31

 ſuch

-1.30

 BEGINNING

-1.27

 teori

-1.27

POSITIVE LOGITS

2.16

↵

1.83

1.39

any

1.38

However

1.37

 about

1.34

いや

1.30

1.26

およ

1.26

Activations Density 0.018%