INDEX

Explanations

explanation:**

This neuron detects the uppercase dateline location headers (city names) that appear at the start of newswire paragraphs.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

</h5>

1.25

<0x0C>

1.20

</img>

1.19

");*/

1.19

</i>

1.18

*/}

1.16

);*/

1.16

""")

1.16

`);

1.14

*/}

1.12

POSITIVE LOGITS

</strong>

1.55

ጹ

0.57

வாகும்

0.54

 મોટી

0.54

ເປັນ

0.53

</b>

0.52

চ্ছন্ন

0.51

 sigur

0.50

。《

0.50

макраты

0.49

Activations Density 0.010%