INDEX

Explanations

percentage rate calculations

The neuron fires on individual content nouns—especially technical or parameter‐type object names (e.g. “logo,” “argument,” “filename,” “prompt,” “sequence,” etc.).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

aptops

0.71

或

0.70

CV

0.66

 eller

0.65

ή

0.65

 certify

0.65

।'

0.62

 roommates

0.61

 ব্যবসা

0.61

或

0.61

POSITIVE LOGITS

 which

0.75

 표현

0.75

 Instead

0.73

which

0.72

 expression

0.71

 expres

0.68

 Expression

0.66

表达

0.65

表達

0.65

 preferring

0.64

Activations Density 0.001%