INDEX

Explanations

bridge between technical and business

The neuron strongly activates on unusually rare or low‐frequency content tokens.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

fax

0.96

 bokeh

0.94

 constituyen

0.91

 tonal

0.89

 tolerable

0.89

 modem

0.87

 soliton

0.87

 fermion

0.87

 diastere

0.87

􀂃

0.87

POSITIVE LOGITS

ck

0.80

age

0.80

cc

0.70

bump

0.69

vc

0.68

halten

0.68

zal

0.67

rand

0.66

ТЕ

0.66

PRODUCTION

0.66

Activations Density 0.000%