INDEX

Explanations

code delimiters (", [, ])

The neuron strongly activates on tokens that are numbers or forward‐slash separators (e.g. version digits, XML tag slashes, and file‐path components).

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

enseignements

-1.35

Spécifications

-1.33

 possibility

-1.24

᳝

-1.20

 aspetto

-1.20

Spécification

-1.20

 rozwią

-1.15

boyfriend

-1.15

Дорогие

-1.15

Apre

-1.13

POSITIVE LOGITS

but

1.37

such

1.29

 SUCH

1.24

1.23

to

1.23

 such

1.21

 Dinas

1.17

 Atención

1.15

 dette

1.14

["

1.10

Activations Density 0.193%