INDEX

Explanations

delta):

This neuron strongly activates on content words that occur immediately before a sentence‐ or clause‐ending punctuation, i.e. on sentence‐final words.

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

.¹

0.88

.³

0.88

 cierta

0.87

 какая

0.86

 осуществля

0.85

 hehe

0.84

 operand

0.83

 какие

0.82

 ciertas

0.82

 gewisse

0.82

POSITIVE LOGITS

ות

0.84

Οι

0.76

栱

0.75

窿

0.72

वहीं

0.70

𝘔

0.69

Kø

0.68

玴

0.68

urow

0.67

Ա

0.65

Activations Density 0.000%