INDEX

Explanations

The neuron activates on numeric tokens and figures (e.g. years, statistics, counts).

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

 stvar

0.55

 bosses

0.54

 veden

0.54

 denomination

0.53

业界

0.53

 underworld

0.52

 preamble

0.52

当局

0.51

ridos

0.51

 jurisdict

0.50

POSITIVE LOGITS

 Assistant

1.63

 assistant

1.62

Assistant

1.62

assistant

1.52

 Analyst

1.48

助理

1.48

 Technician

1.44

 technician

1.43

 Coordinator

1.40

 coordinator

1.38

Activations Density 0.379%