INDEX

Explanations

multi-variable expressions

This neuron fires on two-variable product terms (adjacent variable pairs like “xy,” “uv,” “ac,” “bd,” etc.) in algebraic expressions.

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

為に

-1.27

镲

-1.27

CodeAnalysis

-1.26

様な

-1.25

 cercanas

-1.23

 JOHN

-1.22

ままで

-1.21

lì

-1.19

 insanely

-1.19

の外

-1.18

POSITIVE LOGITS

 Mostly

1.48

 Relatively

1.45

 cautious

1.42

 Surprisingly

1.41

 When

1.41

 Primarily

1.41

 perpetually

1.40

 Perhaps

1.39

 three

1.39

 four

1.38

Activations Density 0.008%