INDEX

Explanations

closing brace then opening brace

This neuron is looking for code‐style tokens—especially programming punctuation (parentheses, braces, operators like “++”) and numeric or code‐specific literals—i.e. bits of source code syntax.

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

now

-1.03

 pacchetto

-0.85

 there

-0.79

 both

-0.79

 their

-0.78

 Crud

-0.77

 some

-0.77

 bahay

-0.75

 almost

-0.75

 various

-0.73

POSITIVE LOGITS

try

0.97

);"

0.96

 Waugh

0.93

>",

0.87

prove

0.86

 JOptionPane

0.85

synchronized

0.85

extend

0.85

}})

0.84

drop

0.83

Activations Density 0.004%