INDEX

Explanations

Pandas or tenant farms

The neuron flags any token containing the substring “pand” (e.g. in “Pand,” “pandas,” “pandera,” etc.).

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

饹

-3.25

 praca

-3.02

ᯅ

-2.97

俛

-2.95

 zoude

-2.91

 veze

-2.89

êmio

-2.88

齪

-2.86



-2.84

嫪

-2.83

POSITIVE LOGITS

when

2.89

2.86

2.83

Have

2.78

 There

2.61

</h3>

2.47

With

2.44

Since

2.42

Although

2.39

Not

2.38

Activations Density 0.007%