INDEX

Explanations

explanation

The neuron strongly activates on direct reader-address phrases of the form “if you’re (or you are) reading this.”

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 pourrez

0.57

 Belgique

0.56

 encuentros

0.55

詣

0.51

dig

0.51

 Scheme

0.51

 visit

0.50

 Lieut

0.50

 séances

0.50

ुलकर

0.50

POSITIVE LOGITS

愚

0.63

Electrical

0.62

Finger

0.61

HTML

0.59

Untitled

0.59

અ

0.59

Reviewed

0.59

әп

0.58

Neither

0.57

之前

0.57

Activations Density 0.000%