INDEX

Explanations

in groups

The neuron strongly activates on large standalone numeric tokens (e.g. multi-digit or floating-point numbers).

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

 which

0.81

which

0.77

 которые

0.75

 entering

0.71

 которых

0.68

 которое

0.68

 WHICH

0.68

 которым

0.67

 который

0.66

 които

0.65

POSITIVE LOGITS

 ausschließlich

0.59

 silos

0.58

这样一个

0.57

 tandem

0.56

 accordance

0.55

:${

0.55

 Darkness

0.55

 ignorance

0.54

 wyłącznie

0.52

middels

0.52

Activations Density 0.353%