INDEX

Explanations

index or init

The neuron is detecting code tokens that refer to file or module path segments—especially occurrences of “index” (e.g. in import/require strings and filenames).

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

-1.42

to

-1.07

for

-1.04

as

-1.03

of

-1.01

cardigan

-0.96

ため

-0.94

 sogenannte

-0.94

 these

-0.93

 into

-0.93

POSITIVE LOGITS

 været

1.19

𝒱

1.08

 vêtement

1.04

dawnictwo

1.01

 ejus

1.00

चार

1.00

Ран

0.99

 Préférences

0.98

PUTY

0.98

්‍ර

0.96

Activations Density 0.015%