INDEX

Explanations

learning rate, recall, batch size

The neuron strongly activates on technical optimization and training terms—like “momentum,” “line search,” “schedule,” “recall,” “learning,” and “batch”—i.e. references to optimization methods or hyperparameter settings.

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

and

-1.96

-1.80

es

-1.78

-1.73

but

-1.64

al

-1.63

om

-1.62

-1.59

-1.56

-1.55

POSITIVE LOGITS

威力

1.53

賂

1.51

 ใหม่

1.45

Bunch

1.42

王妃

1.40

用の

1.39

nim

1.38

 veremos

1.38

 nghiệ

1.38

 использу

1.36

Activations Density 0.001%