INDEX

Explanations

emojis and punctuation

The neuron is tuned to detect subjective commentary or personal‐opinion statements—especially clauses marked by modals, questions, exclamations, and evaluative 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

Negative Logits

 aswell

0.67

 inoltre

0.58

 furthermore

0.55

 unregulated

0.54

.%

0.53

 انها

0.53

 crucially

0.53

 ಸರ್ಕಾರ

0.50

ন্দন

0.50

ー

0.50

POSITIVE LOGITS

;-)

2.86

 😉

2.77

😉

2.70

:-)

2.64

😂

2.64

 😂

2.63

🤣

2.61

;)

2.58

😜

2.56

😁

2.53

Activations Density 0.623%