INDEX

Explanations

store, collection, loader, statistics

The neuron fires on identifiers and keywords tied to data‐handling components—especially “data,” “DataLoader,” “dataset,” “transform,” and related loader/sampler names.

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

-2.02

 about

-1.99

-1.85

-1.60

if

-1.46

 ausdrück

-1.42

ons

-1.41

時候

-1.40

and

-1.39

所有

-1.34

POSITIVE LOGITS

⫺

1.57

didSet

1.50

Ҭ

1.42

 우리

1.40

ття

1.38

ƌ

1.38

Выводы

1.36

pired

1.36

スプレー

1.36

 solide

1.35

Activations Density 0.004%