INDEX

Explanations

risk profiles

This neuron fires most strongly on punctuation used to separate clauses (especially commas) and on the very next token that begins a new phrase or sentence.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

ت

0.93

ین

0.86

Pure

0.76

ight

0.73

la

0.71

Older

0.70

Courtesy

0.70

ם

0.69

If

0.68

ת

0.68

POSITIVE LOGITS

 protegido

0.88

ARAJYA

0.84

 prepd

0.83

🙉

0.82

Khokhlov

0.82

ORES

0.79

 чуть

0.76

crystals

0.76

 xhrObj

0.76

ফটেন্যান্ট

0.75

Activations Density 0.003%