INDEX

Explanations

at + specified points

The neuron never activates on any token—it appears to be effectively “dead,” detecting no particular pattern.

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

was

-2.05

欒

-1.62

…»

-1.62

of

-1.61

ro

-1.61

!*\

-1.54

 الصيف

-1.52

for

-1.52

isasi

-1.51

ización

-1.50

POSITIVE LOGITS

2.97

2.59

you

2.59

the

1.71

1.66

總是

1.63

 carreteras

1.56

 всех

1.55

メロン

1.50

ांकि

1.48

Activations Density 0.087%