INDEX

Explanations

foreign languages

The neuron strongly responds to commas and other high-frequency function words or discourse markers (e.g. “so,” “often,” “I,” “it”), effectively picking out punctuation and common stop-type words in the text.

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

を選択します

-0.96

パッド

-0.92

redacted

-0.87

ഃ

-0.86

Hrsg

-0.86

 fhort

-0.85

韩

-0.84

you

-0.84

sandalias

-0.82

𠮷

-0.82

POSITIVE LOGITS

 approach

0.98

 Muchos

0.95

 versões

0.92

 regiões

0.90

はこの

0.90

 Muchas

0.88

 سيكون

0.88

 vicini

0.86

 response

0.85

 apabila

0.85

Activations Density 0.024%