INDEX

Explanations

thanks and appreciated

The neuron flags polite “help‐me” language—places where the author asks for assistance or expresses appreciation (e.g. “any help is greatly appreciated,” “thank you”).

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

....

-1.05

…..

-0.96

.....

-0.96

….

-0.94

.......

-0.93

 oxigen

-0.93

 skool

-0.93

😒

-0.93

...');

-0.90

...

-0.90

POSITIVE LOGITS

(:

1.07

 hopefully

0.96

:"

0.93

 kayna

0.91

 haha

0.91

 fellow

0.90

also

0.89

 siempre

0.89

aurais

0.89

lysm

0.88

Activations Density 0.035%