INDEX

Explanations

code preparation and actions

The neuron strongly activates on numeric expressions—years, counts, measures, and related punctuation (e.g. question marks around “Planning an event in…?”)—so it’s essentially detecting quantitative references in the text.

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

night

0.78

massive

0.75

together

0.73

 representando

0.72

morning

0.71

organizations

0.71

clothing

0.71

JF

0.71

lots

0.70

明らか

0.70

POSITIVE LOGITS

్

0.80

 méth

0.79

 tím

0.78

še

0.76

 Đường

0.75

 ilgili

0.74

్ర

0.73

 mezi

0.73

më

0.73

 quím

0.73

Activations Density 0.001%