INDEX

Explanations

mathematical conclusions

This neuron picks out structural cue words used to introduce steps or clauses in formal mathematical statements—words like “Let,” “Suppose,” “Then,” “Any,” and “for.”

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

Untuk

-1.55

癒し

-1.45

いきましょう

-1.42

うれ

-1.38

 píše

-1.37

倜

-1.36

んですが

-1.36

to

-1.35

余談

-1.34

 eenvoudig

-1.33

POSITIVE LOGITS

not

1.59

all

1.59

？！”

1.38

how

1.33

in

1.33

 made

1.27

 women

1.25

</strong>

1.22

has

1.21

 не

1.20

Activations Density 0.073%