INDEX

Explanations

have or exist

The neuron fires strongly on verbs and modal auxiliaries that describe actions or capabilities (e.g. “delete,” “create,” “can,” “have,” “will”).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 volverá

-1.72

 BOTH

-1.52

雰囲気が

-1.42

But

-1.39

 αλλά

-1.38

BUT

-1.38

 EVERY

-1.34

 pasará

-1.34

ὁ

-1.34

\&

-1.33

POSITIVE LOGITS

tismo

1.58

ּוֹ

1.56

荣誉

1.54

erobic

1.41

íté

1.40

₁)

1.35

MfG

1.33

ַּ

1.30

<bos>

1.28

ถ้า

1.27

Activations Density 0.069%