INDEX

Explanations

command line commands

The neuron specifically detects the beginning‐of‐sequence (start‐of‐sentence) token.

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

dui

0.39

 persuaded

0.35

na

0.35

 например

0.34

 changed

0.33

 violence

0.33

 Xinhua

0.33

 Zulu

0.33

 several

0.33

 moral

0.32

POSITIVE LOGITS

Си

0.41

Су

0.39

ńskich

0.38

မိ

0.37

Це

0.36

 Су

0.36

قب

0.36

نګ

0.35

 Closeup

0.35

وث

0.35

Activations Density 0.393%