INDEX

Explanations

The neuron fires strongly on standalone comment‐marker tokens or long runs of them (e.g. “#…#” or “//…//”), i.e. it detects comment delimiters or separator lines in code.

code delimiters and comments

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

 své

-1.14

 earlier

-1.11

 early

-0.98

in

-0.98

最早

-0.98

 sufrió

-0.96

escolar

-0.95

 Ländern

-0.93

מית

-0.92

люми

-0.92

POSITIVE LOGITS

 then

1.28

 acabamento

1.18

 płyn

1.14

 another

1.13

ŉ

1.11

 mariana

1.05

 depois

1.02

麻煩

1.00

maxcdn

1.00

Activations Density 0.003%