INDEX

Explanations

study the

The neuron strongly activates on instances of the research‐method verb “study.”

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

驺

-1.66

izzera

-1.52

ITHER

-1.50

adena

-1.45

chsel

-1.45

踯

-1.39

 schwarze

-1.38

 complicado

-1.38

他才

-1.38

ienced

-1.38

POSITIVE LOGITS

 maxn

1.53

of

1.49

 with

1.47

Или

1.45

Размеры

1.42

it

1.39

ньому

1.38

 tapi

1.38

Марина

1.38

Его

1.34

Activations Density 0.151%