INDEX

Explanations

terms and conditions

The neuron fires on words that appear in software‐license or copyright boilerplate (e.g. “the,” “any,” “express,” “implied,” project/author attributions, etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 however

-1.29

命运

-0.81

Si

-0.79

停车

-0.78

 thậm

-0.78

fh

-0.77

cargas

-0.77

 cunos

-0.76

सम

-0.75

ofd

-0.75

POSITIVE LOGITS

Rode

0.84

скур

0.83

 тоже

0.83

\{\

0.83

pires

0.83

aved

0.80

 بينما

0.79

('/');

0.79

 عندما

0.77

ροφο

0.77

Activations Density 0.240%