INDEX

Explanations

according to sources

The neuron consistently activates on tokens involved in presenting numeric statistics or data‐source metadata (e.g. numbers, decimal values, and source labels like “data,” “from,” or agency names) in the text.

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

 手提

-1.35

Shorts

-1.35

Meli

-1.34

枕

-1.33

 sekarang

-1.33

tronlink

-1.31

 потім

-1.30

gründung

-1.29

вні

-1.28

 zato

-1.27

POSITIVE LOGITS

ion

1.52

 this

1.45

1.39

of

1.36

our

1.34

data

1.32

ou

1.31

</b>

1.31

thing

1.30

 that

1.28

Activations Density 0.052%