INDEX

Explanations

want or need

The neuron detects copular or passive constructions—i.e. forms of “to be” (is, was, will be, etc.) linking to an adjective or past participle.

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

of

-1.88

don

-1.18

ции

-1.12

المصادر

-1.08

 said

-1.06

不宜

-1.05

 모두

-1.00

-0.99

се

-0.98

 не

-0.97

POSITIVE LOGITS

it

2.02

而是

1.59

to

1.44

 instead

1.43

 Instead

1.35

 vielmehr

1.35

Instead

1.28

 versátil

1.28

倨

1.22

instead

1.20

Activations Density 0.200%