INDEX

Explanations

mystery and explanation

This neuron fires on semantically content-heavy words (i.e. non–stopwords such as key nouns, verbs, and adjectives) throughout the text.

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

 unrelenting

0.93

 requisite

0.90

 bereit

0.86

 inclement

0.85

 grueling

0.84

 torrential

0.82

 impactful

0.82

 relentless

0.80

 erforder

0.80

 flurry

0.80

POSITIVE LOGITS

 theories

1.34

謎

1.27

解釋

1.27

 Theories

1.25

猜测

1.24

 интерпре

1.23

 explanation

1.21

解释

1.21

 interpretations

1.21

 explanations

1.20

Activations Density 2.288%