INDEX

Explanations

best practices, best way, best for

The neuron fires strongly on comparative and superlative adjectives (e.g. “better,” “best,” “most”) indicating ranking or improvement.

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

0.66

0.64

raut

0.62

0.61

тивным

0.60

দার

0.59

 great

0.59

0.58

 Myths

0.57

 Generations

0.55

POSITIVE LOGITS

 suited

1.54

suited

1.38

 behaved

1.10

sellers

1.02

behaved

0.99

owing

0.93

 geeignet

0.91

equipped

0.91

 versed

0.91

 equipped

0.89

Activations Density 0.403%