INDEX

Explanations

medal, appear, csrf, contribution

The neuron is detecting programming‐language code tokens, particularly import/using statements and library or namespace identifiers.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 Connors

-0.67

 getTarget

-0.66

resser

-0.64

 птица

-0.63

 gained

-0.63

 lunghi

-0.62

Summing

-0.62

дали

-0.61

sourceMap

-0.61

any

-0.61

POSITIVE LOGITS

ཡ

0.79

nities

0.73

gm

0.71

áver

0.70

 Somalia

0.69

ミルク

0.69

Scy

0.66

FETCH

0.66

ikti

0.65

 anvil

0.65

Activations Density 0.072%