INDEX

Explanations

mathematical concepts and equations

The neuron fires on the little “glue” words that set up conditions in math writing—tokens like “in,” “for,” “where,” “and,” “all,” “let,” etc., used to introduce hypotheses or quantifiers.

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

<bos>

-2.42

-2.05

-2.00

-1.95

܇

-1.92

-1.81

存知

-1.80

 Barcelone

-1.79

POSITIVE LOGITS

the

2.27

it

2.03

 these

1.83

 pretende

1.71

遑

1.67

 know

1.66

 したい

1.65

峮

1.63

 posteriormente

1.60

we

1.57

Activations Density 0.034%