INDEX

Explanations

place names and proper nouns

This neuron detects the leading subword of multi-token proper names—i.e. the first token of a capitalized named entity.

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

Only

-1.33

Other

-1.31

Ці

-1.20

Both

-1.18

Sometimes

-1.13

だよね

-1.11

 Only

-1.10

たび

-1.08

Ironically

-1.07

 відповід

-1.05

POSITIVE LOGITS

 dette

1.26

со

1.16

⩕

1.13

Ų

1.10

から

1.09

rr

1.09

 johnson

1.07

.-

1.05

ų

1.05

 durante

1.03

Activations Density 0.216%