INDEX

Explanations

logic, budgets, moments, facility

The neuron is sensitive to longer, more content-heavy (mid-frequency) words—i.e. formal or topic-specific vocabulary rather than common function words.

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Not in Any Lists

Negative Logits

 relating

-1.31

 this

-1.29

 possibly

-1.20

 after

-1.18

one

-1.18

 gives

-1.12

 includes

-1.09

get

-1.09

 Very

-1.09

 numerous

-1.09

POSITIVE LOGITS

 véritables

1.24

 trum

1.19

rute

1.19

hattan

1.17

 dando

1.17

 cuadrado

1.16

 beſt

1.15

vœ

1.14

และ

1.13

 accesorios

1.13

Activations Density 0.028%