INDEX

Explanations

one over others

The main thing this neuron does is spot comparative or preference constructions that contrast one option against another (e.g. “one over the other,” “better,” “more,” “prefer”).

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

>>>>>>>

-1.39

Even

-1.39

‼

-1.38

</u>

-1.36

NetworkInfo

-1.32

<0xCF>

-1.30

ggg

-1.29

in

-1.29

 formar

-1.27

After

-1.24

POSITIVE LOGITS

 aceptación

1.57

</h2>

1.41

jenige

1.36

 cooperación

1.32

 interacción

1.29

 depending

1.29

bueno

1.29

怎样

1.26

 excelencia

1.25

 prevención

1.24

Activations Density 0.072%