INDEX

Explanations

selection of

The main thing this neuron does is spot terms that emphasize large quantities or extensive selections.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

に従

-0.81

ится

-0.80

 commanded

-0.76

Здравствуйте

-0.75

zí

-0.74

äste

-0.73

 различни

-0.73

 commands

-0.72

 degeneration

-0.71

rientes

-0.71

POSITIVE LOGITS

 selection

1.13

 products

1.06

selection

0.89

 Lifestyle

0.86

Online

0.82

yli

0.82

 selections

0.81

kében

0.81

casian

0.81

Selection

0.80

Activations Density 0.005%