INDEX

Explanations

Louisa May Alcott characters

This neuron is tracking a token’s position in the overall document, firing (with strength proportional to activation value) more on later tokens in the input.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 Servers

-0.83

Servers

-0.83

 Commande

-0.78

 lacquer

-0.78

 wanda

-0.78

 رک

-0.77

 cahier

-0.75

webView

-0.75

Retention

-0.74

 Cabinets

-0.74

POSITIVE LOGITS

 Alcott

1.29

Jo

1.24

 Louisa

1.13

 Laurie

1.10

Meg

1.03

 Concord

1.02

 Little

0.99

Meg

0.97

 March

0.95

Jo

0.93

Activations Density 0.019%