INDEX

Explanations

Russian and English words about existence or possibility

The neuron detects words that convey future-oriented predictions or modal emphasis (e.g. “will,” “can,” “be,” “success,” “luck,” “important”).

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

1.04

0.98

0.97

0.90

0.85

0.83

0.82

0.81

0.79

0.78

POSITIVE LOGITS

 dość

0.97

і

0.96

 அவசியம்

0.94

 тільки

0.92

 imprescindible

0.91

 кілька

0.89

 scanty

0.89

 або

0.88

 своє

0.88

 досить

0.88

Activations Density 0.002%