INDEX

Explanations

remastered Bey Boi IANS Ibadan Lange

The neuron strongly activates on capitalized multi‐letter tokens—i.e. proper names, trademarks, titles, or acronyms.

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

</strong>

-3.33

↵

-2.77

</h3>

-2.66

 have

-2.61

-2.55

-2.53

te

-2.48

ti

-2.42

禤

-2.41

桕

-2.39

POSITIVE LOGITS

頑張って

2.39

 komplette

2.36

 mittlere

2.34

苳

2.34

 offizielle

2.23

transQ

2.22

長く

2.22

 bessere

2.19

 entsprechende

2.17

 dochter

2.17

Activations Density 0.033%