INDEX

Explanations

of followed by currency or percentage

The neuron never activates on any token here—it appears to be inactive (i.e. it doesn’t detect any particular feature in these samples).

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

These

-1.84

Some

-1.80

of

-1.78

Like

-1.77

 opérés

-1.76

 with

-1.76

what

-1.73

Even

-1.71

Very

-1.70

Knowing

-1.70

POSITIVE LOGITS

1.94

Ô

1.82

旆

1.80

</i>

1.79

,;

1.75

‚

1.70

泚

1.66

 !!!!

1.63

</b>

1.62

ï

1.61

Activations Density 0.024%