INDEX

Explanations

initial stage

The neuron strongly activates on words denoting the “first” or “beginning” stage (e.g. initial, beginning, introductory).

New Auto-Interp

Configuration

Prompts (Dashboard)

392,802 prompts, 256 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

 перший

0.71

 actualidad

0.69

 prestigious

0.69

 first

0.68

centric

0.66

prest

0.66

ται

0.65

piegel

0.65

 Weitere

0.65

 esteemed

0.64

POSITIVE LOGITS

Initial

1.04

 Initial

1.02

 Initialization

0.96

 initialization

0.94

INITIAL

0.93

Initialization

0.92

 Init

0.89

 stages

0.89

 Inicial

0.89

आती

0.89

Activations Density 0.241%