INDEX

Explanations

recursive, Latin, Monte Carlo

The neuron fires strongly on mid-frequency content words—i.e. nouns and verbs that carry core semantic meaning rather than on common function 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

 різних

0.83

ちょっと

0.77

 messaging

0.76

 portrayal

0.74

SPS

0.74

 preaching

0.72

 verifiable

0.72

 طور

0.72

 setMessage

0.71

 portraying

0.70

POSITIVE LOGITS

ag

0.80

oph

0.80

 angegeben

0.80

ases

0.79

om

0.79

 وبر

0.78

 nær

0.77

ates

0.76

 śred

0.76

ay

0.75

Activations Density 0.000%