INDEX

Explanations

GPT-4 and AI models

The neuron consistently activates on numeric tokens—especially numerical figures and statistical values (including decimals).

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

http

0.65

ẳn

0.65

at

0.63

 দেবার

0.61

 muß

0.59

某种

0.58

mm

0.57

 Exam

0.57

 Stratford

0.57

POSITIVE LOGITS

 العديد

1.11

 países

1.02

 netizens

1.02

 impactful

0.98

 elites

0.97

 metaverse

0.96

 sicuramente

0.94

 frequentemente

0.93

 esports

0.92

 estadounidenses

0.92

Activations Density 0.188%