INDEX

Explanations

program keywords

The neuron strongly activates on capitalized section or article titles and headlines (e.g. multi‐word headings with dashes or colons).

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

ikale

0.61

𝚝

0.59

丁寧に

0.59

 স্থিত

0.58

unta

0.57

 तपाईं

0.56

 إنشاء

0.55

積極的に

0.55

াজে

0.54

েন্টের

0.54

POSITIVE LOGITS

0.60

♑

0.60

 Huffington

0.57

 에서

0.55

 Forbes

0.54

것도

0.54

 volg

0.53

 Boolean

0.52

 Youtube

0.52

 Shang

0.52

Activations Density 0.030%