INDEX

Explanations

can be followed by adjective participle

The neuron consistently fires on the base form of the verb “be” (especially in “can be” or “be done” constructions).

New Auto-Interp

Configuration

Prompts (Dashboard)

24,576 prompts, 128 tokens each

Dataset (Dashboard)

monology/pile-uncopyrighted

Embeds

IFrame

Link

Negative Logits

Ⴈ

-2.03

and

-1.83

就没

-1.66

 ขาย

-1.66

tifications

-1.66

 gleiche

-1.63

"¿

-1.61

tified

-1.60

㍑

-1.58

琑

-1.58

POSITIVE LOGITS

 that

2.47

2.42

 some

1.88

all

1.75

1.70

！】

1.58

ла

1.55

section

1.55

 що

1.52

 consentimento

1.52

Activations Density 0.023%