INDEX

Explanations

The neuron flags uncommon or “rare” subword tokens—especially those coming from proper names, technical terms, or other low‐frequency words.

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

No Comments

Negative Logits

Ка

0.25

 הכ

0.25

Ми

0.24

りました

0.23

Следу

0.22

Setelah

0.22

 Фи

0.22

ίνεται

0.22

짐

0.22

WithError

0.22

POSITIVE LOGITS

 periodontal

0.31

 abnorm

0.27

 kiosks

0.27

 postmodern

0.26

 petrochemical

0.26

specific

0.25

have

0.25

 spécifiques

0.25

 silos

0.25

 celestial

0.25

Activations Density 0.000%