INDEX

Explanations

The neuron reliably flags content‐bearing words—nouns, verbs, and technical terms—that carry the main topic or emphasis of a passage.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

都

0.65

都被

0.64

 ஆகியோர்

0.64

 باشند

0.59

 bhavanti

0.58

 Wants

0.57

fought

0.56

 みんな

0.55

 mają

0.55

存在する

0.54

POSITIVE LOGITS

is

2.42

isn

2.25

 wasn

1.99

是不

1.91

was

1.90

 seems

1.87

 adalah

1.87

是一种

1.84

 betekent

1.83

是很

1.82

Activations Density 1.311%