INDEX

Explanations

quadratic formula

The neuron fires strongly on apostrophe‐marked tokens (e.g. ’s, ’re, ’ll) – in other words, it’s detecting contractions and possessive markers.

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

hips

0.81

 Crypt

0.77

 <!--

0.76

 McDaniel

0.76

 fortifications

0.74

 Antip

0.74

all

0.73

 well

0.73

 Arist

0.73

见

0.73

POSITIVE LOGITS

 demandé

0.86

upsilon

0.86

nicheskij

0.84

 kontaktu

0.83

 sicherlich

0.82

zoek

0.82

attu

0.81

 совета

0.80

 лига

0.80

 apki

0.80

Activations Density 0.001%