INDEX

Explanations

rather than

The neuron responds to numeric‐quantity expressions and measurement/comparison terms (i.e. tokens representing numbers or words like “value,” “range,” “comparisons,” “hard‐coding,” “adding,” “left/right,” “approximately,” etc.).

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

 valamint

-1.44

几分

-1.32

きは

-1.30

 ועוד

-1.27

so

-1.26

 nonché

-1.25

and

-1.23

not

-1.20

 millón

-1.20

 berbagai

-1.18

POSITIVE LOGITS

 ones

1.54

而是

1.48

或

1.47

หรือ

1.39

archiviato

1.38

it

1.35

 resultaat

1.26

contentValues

1.25

りたい

1.20

昨晚

1.20

Activations Density 0.108%