INDEX

Explanations

followed by of course

The neuron is tuned to spot discourse‐marker phrases (e.g. “Of course,”) marked by a following comma.

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

*/

-1.75

"],

-1.64

is

-1.63

"]]

-1.55

 curent

-1.53

ть

-1.50

谖

-1.49

の概要

-1.48

'],

-1.48

 предлагает

-1.46

POSITIVE LOGITS

欵

1.66

co

1.62

1.59

 that

1.52

 time

1.48

 pioneering

1.48

and

1.43

橚

1.43

to

1.42

Activations Density 0.005%