INDEX

Explanations

explaining budgets, skills, contracts, and style

This neuron flags unusually rare or out-of-distribution tokens (e.g. odd names, uncommon compounds, legal citations, brand/model terms, etc.).

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

detalle

0.89

ϱ

0.79

vrd

0.77

จัดการ

0.75

lej

0.74

펴

0.73

]").

0.71

gruppe

0.71

 ansvar

0.71

빱

0.71

POSITIVE LOGITS

فات

0.82

 lush

0.77

غو

0.76

 fitted

0.73

 finish

0.72

 zero

0.70

DOE

0.70

ب

0.69

 could

0.69

ব

0.69

Activations Density 0.001%