INDEX

Explanations

any questions, a responsibility

The neuron strongly fires on descriptive content words—especially evaluative adjectives and intensifiers (e.g. “special,” “huge,” “creative,” “new,” “better”).

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

 blev

0.66

 verilen

0.62

 houve

0.61

 contesto

0.58

 effort

0.57

 avrebbe

0.57

 pudiera

0.55

 zouden

0.55

 pudieran

0.55

 Effort

0.54

POSITIVE LOGITS

 penchant

1.30

 tendency

1.22

 reputation

1.09

 propensity

1.02

 reputations

1.01

 fondness

0.96

 knack

0.96

 tendencies

0.93

 aversion

0.92

 желание

0.89

Activations Density 0.617%