INDEX

Explanations

asked what

The neuron fires on tokens that signal interview‐style questions or Q&A transitions (e.g. “asked,” “what,” “when pressed”).

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

and

-1.37

since

-1.36

Since

-1.27

-1.26

This

-1.24

though

-1.24

年も

-1.19

right

-1.16

 تريد

-1.16

.}

-1.15

POSITIVE LOGITS

часы

1.31

🡺

1.28

Съ

1.23

carros

1.20

鸸

1.16

pergillus

1.15

 horloge

1.14

 nemá

1.14

 phénom

1.13

 motivadoras

1.13

Activations Density 0.045%