INDEX

Explanations

review

The neuron fires strongly on words that name or label document sections or calls‐to‐action—e.g. “Giveaway,” “Review,” “Analysis,” “Directions,” “Problems,” “Responses”—in effect picking out section headers.

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

 thậm

-1.37

-1.34

甚至是

-1.25

わたしは

-1.20

-1.18

ypeł

-1.16

 poprzez

-1.16

场面

-1.14

 certaines

-1.13

 maceta

-1.13

POSITIVE LOGITS

 their

1.46

its

1.43

 both

1.40

сных

1.32

 Vikipedi

1.27

 suoi

1.14

鈮

1.13

gió

1.13

 such

1.11

ਜ

1.11

Activations Density 0.417%