INDEX

Explanations

programming code syntax

The neuron fires on symbol‐ and punctuation-heavy tokens, in particular bits of code or file paths (e.g. regex operators, slashes, brackets, backslashes, carets, etc.).

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

inec

0.55

cares

0.55

सत्ता

0.53

‏:‏

0.52

三次

0.51

ဩ

0.51

processes

0.50

gitignore

0.50

низу

0.50

 высокие

0.50

POSITIVE LOGITS

$_

0.74

"#

0.74

%``

0.73

 そう

0.72

'#

0.71

 อย่าง

0.70

*_

0.69

"-

0.66

`${

0.66

 }{}_{\

0.66

Activations Density 0.173%