INDEX

Explanations

end to

Neuron 4 fires on words signaling termination or cessation—e.g. “end,” “stop,” “halt,” or “kibosh.”

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

 that

-1.80

ichtete

-1.69

umbnail

-1.68

謳

-1.63

 poniendo

-1.59

繆

-1.58

考えると

-1.55

 fairly

-1.53

貴重な

-1.52

ربع

-1.52

POSITIVE LOGITS

2.00

it

1.96

1.91

絰

1.84

ia

1.83

колко

1.80

ip

1.80

}\

1.79

你

1.78

ts

1.75

Activations Density 0.007%