INDEX

Explanations

actions followed by "by" or equivalents

The neuron strongly activates on past‐tense verbs (i.e., words ending in “-ed”).

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

 kendini

1.00

 своем

0.87

 ettiği

0.84

 свою

0.84

limit

0.82

 தன்னை

0.82

recent

0.80

kait

0.79

実は

0.78

けど

0.78

POSITIVE LOGITS

 oleh

1.46

by

1.29

 توسط

1.22

 traités

0.93

โดย

0.90

by

0.90

 treated

0.85

 tarafından

0.84

 Oleh

0.82

BY

0.81

Activations Density 1.133%