INDEX

Explanations

squaring, square root, exponent

The main thing this neuron does is pick out mathematical or numerical content—tokens in formulas, theorem statements, and other equation‐ or number‐heavy passages.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

Undo

0.60

 Undo

0.55

 нужда

0.54

ohi

0.53

alsey

0.51

 rock

0.51

 Boom

0.51

abbo

0.49

 lark

0.49

ande

0.49

POSITIVE LOGITS

 squared

1.73

squared

1.56

²

1.45

 squares

1.44

 Squared

1.41

 squaring

1.41

平方

1.40

²,

1.39

 cuadrados

1.38

Squared

1.38

Activations Density 0.051%