INDEX

Explanations

in its design

This neuron detects prominent subjective evaluative adjectives and adverbs (intensifiers) that convey strong quality judgments (e.g. “buzzy,” “reflective,” “incredibly,” “incredible,” “great”).

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

 solchen

0.79

 такого

0.71

разу

0.68

 دیگری

0.66

そのような

0.64

 referido

0.63

 solche

0.62

dą

0.62

 unsuccessful

0.62

 مشروع

0.61

POSITIVE LOGITS

and

0.82

but

0.75

er

0.68

all

0.68

and

0.68

ですが

0.67

です

0.66

而且

0.66

all

0.65

Activations Density 0.088%