INDEX

Explanations

program structure and conditions

The neuron fires on numeric tokens (literals like “63”, “36.5”, etc.) and comparison operators (e.g. “>”) in code snippets.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

-1.41

 ebenfalls

-1.34

“

-1.30

↵↵

-1.30

zelfde

-1.27

几

-1.24

zig

-1.23

 også

-1.22

ANDS

-1.21

MAK

-1.20

POSITIVE LOGITS

 this

1.61

huevos

1.60

</em>

1.55

>−

1.50

italien

1.47

</sup>

1.34

中略

1.33

</strong>

1.33

 (!__

1.31

werkers

1.30

Activations Density 0.005%