INDEX

Explanations

model design

The neuron activates strongly on principal entity nouns—especially names of organizations, sports terms, or topic‐defining subjects (e.g. “team,” “game,” “business,” “club,” “track,” “site,” etc.).

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

 Российской

0.61

larla

0.60

ների

0.58

otten

0.58

喜

0.58

uploads

0.56

TargetPath

0.55

 Их

0.55

ációs

0.55

sums

0.55

POSITIVE LOGITS

wide

1.07

 wide

0.98

mates

0.86

 முழுவதும்

0.84

 motto

0.82

全体の

0.77

 manager

0.75

 keseluruhan

0.73

Wide

0.71

 oficialmente

0.70

Activations Density 0.685%