INDEX

Explanations

replace

The neuron is triggered by occurrences of the word “replace.”

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

是用

-2.34

}$,

-2.14

賤

-2.08

 travels

-2.03

to

-2.02

 една

-2.02

 asesin

-2.00

↵↵

-1.98

fecture

-1.95

 accesses

-1.92

POSITIVE LOGITS

↵

3.75

También

2.23

the

2.19

which

2.03

 Altogether

1.99

 vivido

1.98

にほんブログ村

1.95

 наших

1.91

脑袋

1.90

或者

1.89

Activations Density 0.019%