INDEX

Explanations

details, attract, attention, resize, tree, guards, key

The neuron fires on UI/metadata tokens—things like navigation labels (“home,” “site,” “menu”), status keywords (“processing,” “viewed”), and other interface‐style or section‐heading words.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

</h6>

0.84

ب

0.79

</em>

0.75

б

0.75

カ

0.75

ВА

0.75

<h6>

0.75

 травня

0.74

♧

0.73

0.71

POSITIVE LOGITS

यें

0.84

ëren

0.82

หมือน

0.80

勍

0.79

đi

0.78

 Grün

0.75

ລິ

0.75

Đi

0.74

тый

0.73

linewidth

0.73

Activations Density 0.000%