INDEX

Explanations

conjunctions and transitional phrases

The neuron fires on open‐class “content” words—nouns, main verbs, adjectives, and adverbs that carry the core meaning—while ignoring common function words.

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

 thanks

-1.00

薨

-0.95

 pData

-0.95

 thofe

-0.94

ContentAsync

-0.93

}$.

-0.93

or

-0.93

 Мексика

-0.91

']:

-0.88

 Reprint

-0.88

POSITIVE LOGITS

 garagem

1.37

 naturally

1.16

有很多

1.16

you

1.13

ごま

1.06

 natürlich

1.05

 prostor

1.05

嘻嘻

1.05

 transferência

1.04

 necessarily

1.03

Activations Density 0.037%