INDEX

Explanations

panoramic views or decelerating

The neuron selectively fires on unusually long, polysyllabic words (including multi-syllable names).

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

</em>

-2.61

搁

-2.47

殽

-2.36

 ficaram

-2.25

ついで

-2.22

緋

-2.20

驁

-2.20

办

-2.20

these

-2.19

 Produkten

-2.19

POSITIVE LOGITS

2.97

 WITH

2.92

</i>

2.86

2.78

真的是

2.77

<td>

2.75

</h1>

2.67

2.55

只是

2.53

que

2.48

Activations Density 0.007%