INDEX

Explanations

handwriting and image recognition

The neuron strongly activates on tokens naming or describing machine-learning datasets, tasks, or benchmarks (e.g. “QM8,” “handwritten,” “digit,” “recognition,” “draw”).

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

 ellos

-1.48

 этого

-1.48

脔

-1.38

zość

-1.35

 They

-1.34

🫃

-1.33

 recomendable

-1.32

encing

-1.31

 Página

-1.31

nąć

-1.30

POSITIVE LOGITS

1.84

to

1.45

 meistens

1.42

 being

1.30

by

1.29

bim

1.27

 häufigsten

1.24

 ところ

1.23

 häufig

1.19

ものである

1.14

Activations Density 0.015%