INDEX

Explanations

programming code structure

The neuron fires on tokens that introduce top‐level code declarations or metadata lines—things like package statements, preprocessor defines, namespace declarations, or diff hunk markers.

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

it

-1.26

any

-1.24

 those

-1.23

 already

-1.20

 both

-1.16

резать

-1.13

 dernière

-1.08

 even

-1.07

 truly

-1.05

 then

-1.02

POSITIVE LOGITS

'',

1.24

'';

1.16

⅔

1.10

 stronghold

1.07

⅓

1.01

的很

0.99

 será

0.98

zeba

0.98

 биз

0.94

や

0.94

Activations Density 0.002%