INDEX

Explanations

The neuron strongly activates on the word “read” (usually as part of a “Read …” call‐to‐action or link prompt).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

ز

1.81

ک

1.59

א

1.57

ק

1.49

אם

1.45

ی

1.43

ра

1.41

kan

1.41

ہ

1.41

ма

1.35

POSITIVE LOGITS

 kepercayaan

1.27

ership

1.26

病情

1.25

 команды

1.17

 prostat

1.17

 neuf

1.16

filesystem

1.16

 जबकि

1.16

 seule

1.15

 herring

1.15

Activations Density 0.081%