INDEX

Explanations

improvements to existing things

The neuron detects occurrences of the word “improvement” or “improvements” (typically in patent section headings).

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

久しぶりの

-1.71

没办法

-1.63

愛知県

-1.56

昨年

-1.55

its

-1.52

因為

-1.52

みなさま

-1.51

 необходи

-1.49

$.

-1.48

POSITIVE LOGITS

 While

1.84

 They

1.80

AIRES

1.62

瑷

1.61

 Also

1.58

Они

1.57

 ремон

1.57

นน

1.52

 oiseau

1.52

 terlalu

1.50

Activations Density 0.025%