INDEX

Explanations

starts of lists or phrases

The main thing this neuron does is detect sequences of dance‐step or footwork instructions (e.g., “step right to side, step left behind, step right cross”).

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

ִּ

-1.78

ּוֹ

-1.77

ֶּ

-1.72

ַּ

-1.59

וּ

-1.52

 atardecer

-1.30

 This

-1.29

ֵּ

-1.27

jadikan

-1.21

ведения

-1.20

POSITIVE LOGITS

ׇ

1.64

 again

1.41

ὴν

1.30

 گیاه

1.27

 continue

1.25

kozások

1.20

继续

1.18

Beantwoorden

1.16

や

1.16

,’

1.13

Activations Density 0.004%