INDEX

Explanations

future intentions and positive opinions

This neuron flags enthusiastic first-person testimonial language—especially “I” or “We” statements often accompanied by exclamation marks.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 fatale

-1.01

ddagger

-0.98

 също

-0.97

liegende

-0.94

 カジュアル

-0.92

⛑

-0.89

malige

-0.89

 jaune

-0.88

Alternatively

-0.88

pê

-0.85

POSITIVE LOGITS

 will

4.91

 would

3.56

 sẽ

2.30

 Will

1.98

 будет

1.92

 θα

1.90

Will

1.87

 definitely

1.80

 będzie

1.79

will

1.77

Activations Density 0.006%