INDEX

Explanations

paper called

This neuron detects relatively rare, multisyllabic content words (uncommon or formal vocabulary tokens).

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

ichni

0.83

acariy

0.83

avkh

0.80

ungkinkan

0.79

ijek

0.78

setKeepCallback

0.77

uillez

0.75

仡

0.73

 recharging

0.73

ifestyle

0.72

POSITIVE LOGITS

ר

0.94

א

0.84

in

0.77

"":

0.75

 announce

0.75

ס

0.75

 comunale

0.71

ג

0.71

נ

0.70

Activations Density 0.000%