INDEX

Explanations

HTML tags

The neuron responds to markup, code, and formatting tokens (e.g. HTML/CSS tags and attributes, LaTeX commands, chemical‐equation keywords).

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

 作者

-0.84

店家

-0.81

ederos

-0.81

紘

-0.79

ÓW

-0.77

悍

-0.75

商家

-0.75

 From

-0.75

原生

-0.75

 muualla

-0.74

POSITIVE LOGITS

</strong>

0.98

.&

0.96

,**

0.94

:</

0.90

,*

0.89

~~~

0.89

(#

0.86

plication

0.86

RootElement

0.86

笞

0.85

Activations Density 0.000%