INDEX

Explanations

patent document references

The neuron fires on numeric tokens, particularly citation numbers, patent/document references, page and volume numbers, and other standalone digits.

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

stery

-1.07

for

-1.06

|--

-0.97

＝

-0.95

 whofe

-0.92

ям

-0.91

윗

-0.89

 утеп

-0.89

ká

-0.88

 домо

-0.88

POSITIVE LOGITS

criar

1.29

 utilizzo

1.19

 móveis

1.15

両

1.12

阯

1.10

 中身

1.09

スポンサーリンク

1.08

 ottima

1.08

 meditación

1.08

 hausse

1.05

Activations Density 0.062%