INDEX

Explanations

A: or A followed by word

The neuron detects the “A:” answer label at the start of an answer.

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

-2.56

</h3>

-2.06

 were

-2.02

),

-1.88

↵

-1.76

-1.59

 każ

-1.59

and

-1.55

-1.48

for

-1.41

POSITIVE LOGITS

 haine

1.89

 déliv

1.88

萏

1.87

硖

1.86

ться

1.78

 construit

1.74

bluza

1.69

zumaki

1.68

 cheia

1.65

knię

1.63

Activations Density 0.028%