INDEX

Explanations

know if

The neuron strongly activates on tokens that begin new sentences (i.e. sentence-initial words and related leading punctuation).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

icia

0.64

্ধ্য

0.62

これからも

0.59

 Embora

0.58

 cumul

0.57

 대통령

0.57

 detectar

0.57

 erythro

0.56

ंसी

0.56

 irreparable

0.56

POSITIVE LOGITS

 оказалось

1.81

 ternyata

1.77

Turns

1.69

 Turns

1.58

果然

1.48

 оказался

1.44

 оказывается

1.42

 turns

1.41

 оказалась

1.40

 okaza

1.39

Activations Density 0.672%