INDEX

Explanations

how things grow or are made

The neuron fires on the main action verb in “How…” questions—i.e. the instructional or question‐form verbs (open, work/works, expose, earn, etc.) that follow “How.”

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

を有する

0.88

を行う

0.86

REGIUNE

0.82

බා

0.81

ヤモンド

0.80

 théorie

0.80

 trzeba

0.78

を発

0.78

 menyebut

0.77

 পাকিস্তান

0.77

POSITIVE LOGITS

↵

0.92

0.82

0.80

0.78

by

0.77

 during

0.77

↵↵

0.77

0.75

.,

0.74

0.72

Activations Density 0.000%