INDEX

Explanations

value and benefit

The neuron is highly activated by occurrences of the word “value,” spotting mentions of value/worth in the text.

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



-2.27

翆

-2.23

llavero

-1.90

colgante

-1.81

饧

-1.65

𖣔

-1.64

皤

-1.64

檮

-1.60

卣

-1.59

alties

-1.57

POSITIVE LOGITS

1.83

While

1.57

See

1.52

was

1.50

在他

1.49

却不

1.48

 समीक्षा

1.48

................

1.45

These

1.45

夲

1.44

Activations Density 0.005%