INDEX

Explanations

history of topics

The neuron fires on prominent content words in section headings or titles (e.g. “History,” “Corruption,” “Life,” “Sciences,” “Valentine,” “Architecture”)—essentially marking the key topical term of a heading.

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

-1.52

 would

-1.41

-1.32

-1.27

(),

-1.25

 have

-1.23

-1.19

or

-1.15

 should

-1.15

POSITIVE LOGITS

ulement

1.13

asdf

1.09

 yuri

1.09

@@@@@

1.07

 часу

1.07

紲

1.06

睏

1.06

ורך

1.05

 suporta

1.05

 längere

1.03

Activations Density 0.146%