INDEX

Explanations

transactions and grammar rules

The neuron spikes on uncommon, content-bearing words (i.e. low-frequency or topic-specific terms) rather than on common function 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

 تمامی

0.93

 niemals

0.91

非常的

0.87

 aquellas

0.83

 sicherlich

0.82

 somente

0.77

 aquella

0.76

 সকল

0.74

 aquellos

0.73

しっかりと

0.73

POSITIVE LOGITS

 partly

0.82

 complicating

0.78

 reconciling

0.77

 interdependent

0.73

 combinatorial

0.72

 variational

0.69

 deftly

0.68

 ambiguities

0.67

 üç

0.66

 nonconvex

0.66

Activations Density 0.044%