INDEX

Explanations

x, y coordinates

This neuron is most strongly activated by occurrences of the variable name “x” in code.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

ando

-1.46

in

-1.39

-1.35

are

-1.34

lós

-1.32

there

-1.29

还有

-1.28

-1.27

では

-1.20

and

-1.20

POSITIVE LOGITS

ecznie

1.35

dersfield

1.34

 marquées

1.32

rativa

1.30

に限

1.30

に似

1.27

景

1.27

exercice

1.23

 trochu

1.18

camisas

1.17

Activations Density 0.008%