INDEX

Explanations

introduces clauses

This neuron activates on clause-linking discourse markers (subordinating or connective adverbs) such as “whereby,” “being,” “where,” the Russian “чём,” the Chinese “并” or “而,” etc.

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

<td>

-1.91

 exorbit

-1.78

]}"

-1.70

 الجديدة

-1.59

则是

-1.59

"@/

-1.59

ಸ್

-1.56

$};

-1.56

クシー

-1.55

 hella

-1.55

POSITIVE LOGITS

 that

2.02

この日

1.57

余談

1.56

 each

1.55

 таким

1.53

ガリ

1.48

そも

1.47

ヤバ

1.47

サード

1.44

 they

1.42

Activations Density 0.011%