INDEX

Explanations

The neuron fires on uppercase “speaker‐label” tokens (e.g. initials or names in all caps marking who’s speaking in an interview transcript).

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

 flames

0.65

 perintah

0.61

 mares

0.57

 sportive

0.57

 fancy

0.56

 caps

0.55

iar

0.55

poi

0.54

 wiecz

0.54

bbq

0.54

POSITIVE LOGITS

 Recruiting

0.69

Integrated

0.67

 Integrated

0.65

Publications

0.63

 Promoting

0.63

 László

0.63

ꯔ

0.63

IA

0.62

Sure

0.62

HK

0.61

Activations Density 0.017%