INDEX

Explanations

training data and code explanation

This neuron activates most strongly on standalone headings, titles, or other section‐level labels (i.e. capitalized or emphasis words marking the start of a new section).

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

my

-1.29

 árboles

-1.27

}:{

-1.26

˂

-1.25

 denominado

-1.24

+*

-1.23

Quais

-1.23

Ainda

-1.21

 SHRI

-1.20

ÑOS

-1.19

POSITIVE LOGITS

></

1.62

 also

1.39

我才

1.23

</tr>

1.23

畀

1.22

椟

1.20

 cioc

1.19

 placed

1.19

勧め

1.18

PDATE

1.17

Activations Density 0.213%