INDEX

Explanations

for a audience

This neuron is primarily detecting high-frequency function words and punctuation (e.g., commas and short “stop” words like “of,” “by,” “it,” “those”).

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

 famosos

0.66

 entrev

0.64

 martyrs

0.59

 muertos

0.59

 vrouw

0.59

 ghosts

0.59

 murdering

0.56

쩌

0.56

 герои

0.55

 quartet

0.55

POSITIVE LOGITS

 practitioners

1.52

 clinicians

1.28

 Practitioners

1.27

 practition

1.27

 practitioner

1.23

 policymakers

1.16

 professionals

1.10

 Practitioner

1.10

 educators

1.08

 readers

1.07

Activations Density 0.191%