INDEX

Explanations

paid / unpaid categorisation

The neuron strongly activates on first-person, personal testimonial language (e.g. “I,” “my,” “helped me,” “could not be happier”), flagging user-style reviews and personal experience statements.

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

 flavorful

1.59

 regimen

1.41

 matchup

1.34

 toward

1.33

 regimens

1.33

 matchups

1.31

 demeanor

1.30

 outfitted

1.29

 geologic

1.27

 nonprofit

1.26

POSITIVE LOGITS

Whilst

1.94

 enquiries

1.86

 dodgy

1.84

 adverts

1.77

 Whilst

1.73

 motorway

1.73

 underpinned

1.73

 HMRC

1.70

 postcode

1.67

 Maths

1.64

Activations Density 2.543%