INDEX

Explanations

Intent extras, keys, or actions

The neuron fires on all-caps identifiers, especially constant or key names in SCREAMING_SNAKE_CASE (e.g. Intent.ACTION_…, Intent.EXTRA_…, database KEY_… tokens).

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

 citada

-1.84

 condicionado

-1.84

silueta

-1.70

 entendido

-1.65

 tomado

-1.60

 with

-1.56

 comprado

-1.52

姈

-1.51

 permitido

-1.49

廕

-1.49

POSITIVE LOGITS

 donc

1.44

which

1.34

for

1.34

 трево

1.34

mathcal

1.33

خدمات

1.31

</i>

1.30

Value

1.30

$\

1.29

guió

1.27

Activations Density 0.004%