INDEX

Explanations

common punctuation and keywords

The neuron fires on tokens that are made up largely of punctuation or symbol sequences (e.g. regex metacharacters, URL pieces, HTML entities, file-path or license-block punctuation).

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

🤑

-0.85

ྀ

-0.84

nue

-0.80

🍉

-0.79

 Informações

-0.79

 Mujeres

-0.79

æa

-0.78

ешься

-0.77

 newInstance

-0.76

 basah

-0.76

POSITIVE LOGITS

onomies

0.80

徙

0.77

Imaginary

0.76

 Nema

0.73

OKAY

0.73

Baptiste

0.72

Gera

0.71

랩

0.71

 bend

0.71

FAB

0.70

Activations Density 0.005%