INDEX

Explanations

references to previous text

The neuron specializes in spotting backward‐reference language in scholarly text—particularly words like “above” and “cited.”

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

 about

-1.38

to

-1.37

 largely

-1.30

 then

-1.27

 around

-1.25

敤

-1.13

 between

-1.13

 along

-1.13

 such

-1.12

POSITIVE LOGITS

 bár

1.20

 pantolon

1.20

👁

1.16

 arbeid

1.15

 Commentaires

1.15

 ucran

1.14

 verdo

1.13

 salvajes

1.13

 parfü

1.13

 fér

1.12

Activations Density 0.001%