INDEX

Explanations

code syntax and reversal

The neuron fires on programmer-defined identifiers (variable names) in code.

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

 first

-0.98

ább

-0.86

 yük

-0.74

 første

-0.73

 ersten

-0.71

 gäller

-0.71

kul

-0.71

 erstmals

-0.71

first

-0.71

chestra

-0.69

POSITIVE LOGITS

[-

3.16

[-

2.28

][-

2.03

(-

1.94

(-

1.89

最後

1.80

最后

1.70

$(-

1.66

 cuối

1.60

$[-

1.59

Activations Density 0.018%