INDEX

Explanations

asking for information politely

The neuron is specialized to detect the modal-conditional phrase “would be.”

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

當

0.90

 когато

0.88

 lorsque

0.85

 Sometimes

0.80

당

0.80

 parfois

0.80

 đã

0.79

 когда

0.79

 иногда

0.79

 cuando

0.78

POSITIVE LOGITS

benefit

0.74

 benefit

0.70

 fain

0.67

ANDE

0.63

gin

0.62

であろう

0.61

 مكون

0.61

受益

0.60

 Benefit

0.59

Benefit

0.59

Activations Density 0.212%