INDEX

Explanations

justify or compensate

This neuron activates on numeric values (integers or decimals) in the text.

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

Glück

-0.87

ība

-0.87

 النساء

-0.82

 both

-0.81

▉

-0.79

Einfach

-0.77

挑戰

-0.77

极为

-0.76

 specifiche

-0.76

イルミネ

-0.75

POSITIVE LOGITS

 justification

1.16

 wenigstens

1.16

justified

1.06

 justified

1.05

Jus

1.03

 compensation

1.02

 justify

1.02

 compensate

1.00

 compensating

0.99

 justifiable

0.97

Activations Density 0.042%