INDEX

Explanations

FOR DEFENDANT-APPELLANT

The neuron detects multi‐token proper nouns and named entities (e.g. place names, personal names, organizations).

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

 some

-0.86

 Large

-0.77

 المل

-0.74

 étant

-0.74

тков

-0.72

avax

-0.69

 mentioned

-0.69

tocin

-0.69

IPA

-0.69

Fetched

-0.68

POSITIVE LOGITS

 elevations

0.83

%\

0.79

 Heli

0.77

erkek

0.75

agaimana

0.74

 truffle

0.74

 protectors

0.73

Hau

0.73

 protégé

0.72

Vall

0.72

Activations Density 0.016%