INDEX

Explanations

punctuation and line breaks

The neuron strongly activates on numeric tokens (digits, numbers, dates, measurements) in the text.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

↵↵

1.30

↵↵↵↵↵↵↵↵↵↵

1.09

↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.05

↵↵↵↵↵↵↵↵

1.04

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.02

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

1.02

↵↵↵↵↵↵↵↵↵↵↵↵

0.98

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.96

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.91

↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵↵

0.88

POSITIVE LOGITS

<0x0D>

0.93

</h2>

0.91

}\

0.84

)\

0.82

↵↵↵

0.75

}}\

0.74

]\

0.74

 […]

0.68

}}\

0.68

))\

0.66

Activations Density 0.021%